import os
import os.path
-import sha
import time
-import tempfile
import re
import platform
import logging
import copy
-import random
from ganeti import ssh
from ganeti import utils
from ganeti import locking
from ganeti import constants
from ganeti import objects
-from ganeti import opcodes
from ganeti import serializer
from ganeti import ssconf
Note that all commands require root permissions.
+ @ivar dry_run_result: the value (if any) that will be returned to the caller
+ in dry-run mode (signalled by opcode dry_run parameter)
+
"""
HPATH = None
HTYPE = None
def __init__(self, processor, op, context, rpc):
"""Constructor for LogicalUnit.
- This needs to be overriden in derived classes in order to check op
+ This needs to be overridden in derived classes in order to check op
validity.
"""
# Dicts used to declare locking needs to mcpu
self.needed_locks = None
self.acquired_locks = {}
- self.share_locks = dict(((i, 0) for i in locking.LEVELS))
+ self.share_locks = dict.fromkeys(locking.LEVELS, 0)
self.add_locks = {}
self.remove_locks = {}
# Used to force good behavior when calling helper functions
# logging
self.LogWarning = processor.LogWarning
self.LogInfo = processor.LogInfo
+ self.LogStep = processor.LogStep
+ # support for dry-run
+ self.dry_run_result = None
for attr_name in self._OP_REQP:
attr_val = getattr(op, attr_name, None)
CheckPrereq, doing these separate is better because:
- ExpandNames is left as as purely a lock-related function
- - CheckPrereq is run after we have aquired locks (and possible
+ - CheckPrereq is run after we have acquired locks (and possible
waited for them)
The function is allowed to change the self.op attribute so that
wanted.append(instance)
else:
- wanted = lu.cfg.GetInstanceList()
- return utils.NiceSort(wanted)
+ wanted = utils.NiceSort(lu.cfg.GetInstanceList())
+ return wanted
def _CheckOutputFields(static, dynamic, selected):
@param lu: the LU on behalf of which we make the check
@param node: the node to check
- @raise errors.OpPrereqError: if the nodes is offline
+ @raise errors.OpPrereqError: if the node is offline
"""
if lu.cfg.GetNodeInfo(node).offline:
raise errors.OpPrereqError("Can't use offline node %s" % node)
+def _CheckNodeNotDrained(lu, node):
+ """Ensure that a given node is not drained.
+
+ @param lu: the LU on behalf of which we make the check
+ @param node: the node to check
+ @raise errors.OpPrereqError: if the node is drained
+
+ """
+ if lu.cfg.GetNodeInfo(node).drained:
+ raise errors.OpPrereqError("Can't use drained node %s" % node)
+
+
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
- memory, vcpus, nics):
+ memory, vcpus, nics, disk_template, disks,
+ bep, hvp, hypervisor_name):
"""Builds instance related env variables for hooks
This builds the hook environment from individual variables.
@param secondary_nodes: list of secondary nodes as strings
@type os_type: string
@param os_type: the name of the instance's OS
- @type status: string
- @param status: the desired status of the instances
+ @type status: boolean
+ @param status: the should_run status of the instance
@type memory: string
@param memory: the memory size of the instance
@type vcpus: string
@param vcpus: the count of VCPUs the instance has
@type nics: list
- @param nics: list of tuples (ip, bridge, mac) representing
- the NICs the instance has
+ @param nics: list of tuples (ip, mac, mode, link) representing
+ the NICs the instance has
+ @type disk_template: string
+ @param disk_template: the disk template of the instance
+ @type disks: list
+ @param disks: the list of (size, mode) pairs
+ @type bep: dict
+ @param bep: the backend parameters for the instance
+ @type hvp: dict
+ @param hvp: the hypervisor parameters for the instance
+ @type hypervisor_name: string
+ @param hypervisor_name: the hypervisor for the instance
@rtype: dict
@return: the hook environment for this instance
"""
+ if status:
+ str_status = "up"
+ else:
+ str_status = "down"
env = {
"OP_TARGET": name,
"INSTANCE_NAME": name,
"INSTANCE_PRIMARY": primary_node,
"INSTANCE_SECONDARIES": " ".join(secondary_nodes),
"INSTANCE_OS_TYPE": os_type,
- "INSTANCE_STATUS": status,
+ "INSTANCE_STATUS": str_status,
"INSTANCE_MEMORY": memory,
"INSTANCE_VCPUS": vcpus,
+ "INSTANCE_DISK_TEMPLATE": disk_template,
+ "INSTANCE_HYPERVISOR": hypervisor_name,
}
if nics:
nic_count = len(nics)
- for idx, (ip, bridge, mac) in enumerate(nics):
+ for idx, (ip, mac, mode, link) in enumerate(nics):
if ip is None:
ip = ""
env["INSTANCE_NIC%d_IP" % idx] = ip
- env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
- env["INSTANCE_NIC%d_HWADDR" % idx] = mac
+ env["INSTANCE_NIC%d_MAC" % idx] = mac
+ env["INSTANCE_NIC%d_MODE" % idx] = mode
+ env["INSTANCE_NIC%d_LINK" % idx] = link
+ if mode == constants.NIC_MODE_BRIDGED:
+ env["INSTANCE_NIC%d_BRIDGE" % idx] = link
else:
nic_count = 0
env["INSTANCE_NIC_COUNT"] = nic_count
+ if disks:
+ disk_count = len(disks)
+ for idx, (size, mode) in enumerate(disks):
+ env["INSTANCE_DISK%d_SIZE" % idx] = size
+ env["INSTANCE_DISK%d_MODE" % idx] = mode
+ else:
+ disk_count = 0
+
+ env["INSTANCE_DISK_COUNT"] = disk_count
+
+ for source, kind in [(bep, "BE"), (hvp, "HV")]:
+ for key, value in source.items():
+ env["INSTANCE_%s_%s" % (kind, key)] = value
+
return env
+def _NICListToTuple(lu, nics):
+ """Build a list of nic information tuples.
+
+ This list is suitable to be passed to _BuildInstanceHookEnv or as a return
+ value in LUQueryInstanceData.
+
+ @type lu: L{LogicalUnit}
+ @param lu: the logical unit on whose behalf we execute
+ @type nics: list of L{objects.NIC}
+ @param nics: list of nics to convert to hooks tuples
+
+ """
+ hooks_nics = []
+ c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
+ for nic in nics:
+ ip = nic.ip
+ mac = nic.mac
+ filled_params = objects.FillDict(c_nicparams, nic.nicparams)
+ mode = filled_params[constants.NIC_MODE]
+ link = filled_params[constants.NIC_LINK]
+ hooks_nics.append((ip, mac, mode, link))
+ return hooks_nics
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
"""Builds instance related env variables for hooks from an object.
@return: the hook environment dictionary
"""
- bep = lu.cfg.GetClusterInfo().FillBE(instance)
+ cluster = lu.cfg.GetClusterInfo()
+ bep = cluster.FillBE(instance)
+ hvp = cluster.FillHV(instance)
args = {
'name': instance.name,
'primary_node': instance.primary_node,
'secondary_nodes': instance.secondary_nodes,
'os_type': instance.os,
- 'status': instance.os,
+ 'status': instance.admin_up,
'memory': bep[constants.BE_MEMORY],
'vcpus': bep[constants.BE_VCPUS],
- 'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
+ 'nics': _NICListToTuple(lu, instance.nics),
+ 'disk_template': instance.disk_template,
+ 'disks': [(disk.size, disk.mode) for disk in instance.disks],
+ 'bep': bep,
+ 'hvp': hvp,
+ 'hypervisor_name': instance.hypervisor,
}
if override:
args.update(override)
(mc_now, mc_max))
-def _CheckInstanceBridgesExist(lu, instance):
+def _CheckNicsBridgesExist(lu, target_nics, target_node,
+ profile=constants.PP_DEFAULT):
+ """Check that the brigdes needed by a list of nics exist.
+
+ """
+ c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
+ paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
+ for nic in target_nics]
+ brlist = [params[constants.NIC_LINK] for params in paramslist
+ if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
+ if brlist:
+ result = lu.rpc.call_bridges_exist(target_node, brlist)
+ result.Raise("Error checking bridges on destination node '%s'" %
+ target_node, prereq=True)
+
+
+def _CheckInstanceBridgesExist(lu, instance, node=None):
"""Check that the brigdes needed by an instance exist.
"""
- # check bridges existance
- brlist = [nic.bridge for nic in instance.nics]
- result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("One or more target bridges %s does not"
- " exist on destination node '%s'" %
- (brlist, instance.primary_node))
+ if node is None:
+ node = instance.primary_node
+ _CheckNicsBridgesExist(lu, instance.nics, node)
class LUDestroyCluster(NoHooksLU):
This checks whether the cluster is empty.
- Any errors are signalled by raising errors.OpPrereqError.
+ Any errors are signaled by raising errors.OpPrereqError.
"""
master = self.cfg.GetMasterNode()
"""
master = self.cfg.GetMasterNode()
result = self.rpc.call_node_stop_master(master, False)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Could not disable the master role")
+ result.Raise("Could not disable the master role")
priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
utils.CreateBackup(priv_key)
utils.CreateBackup(pub_key)
locking.LEVEL_NODE: locking.ALL_SET,
locking.LEVEL_INSTANCE: locking.ALL_SET,
}
- self.share_locks = dict(((i, 1) for i in locking.LEVELS))
+ self.share_locks = dict.fromkeys(locking.LEVELS, 1)
def _VerifyNode(self, nodeinfo, file_list, local_cksum,
- node_result, feedback_fn, master_files):
+ node_result, feedback_fn, master_files,
+ drbd_map, vg_name):
"""Run multiple tests against a node.
Test list:
- compares ganeti version
- - checks vg existance and size > 20G
+ - checks vg existence and size > 20G
- checks config file checksum
- checks ssh to other nodes
@param node_result: the results from the node
@param feedback_fn: function used to accumulate results
@param master_files: list of files that only masters should have
+ @param drbd_map: the useddrbd minors for this node, in
+ form of minor: (instance, must_exist) which correspond to instances
+ and their running status
+ @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
"""
node = nodeinfo.name
# compares ganeti version
local_version = constants.PROTOCOL_VERSION
remote_version = node_result.get('version', None)
- if not remote_version:
+ if not (remote_version and isinstance(remote_version, (list, tuple)) and
+ len(remote_version) == 2):
feedback_fn(" - ERROR: connection to %s failed" % (node))
return True
- if local_version != remote_version:
- feedback_fn(" - ERROR: sw version mismatch: master %s, node(%s) %s" %
- (local_version, node, remote_version))
+ if local_version != remote_version[0]:
+ feedback_fn(" - ERROR: incompatible protocol versions: master %s,"
+ " node %s %s" % (local_version, node, remote_version[0]))
return True
- # checks vg existance and size > 20G
+ # node seems compatible, we can actually try to look into its results
bad = False
- vglist = node_result.get(constants.NV_VGLIST, None)
- if not vglist:
- feedback_fn(" - ERROR: unable to check volume groups on node %s." %
- (node,))
- bad = True
- else:
- vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
- constants.MIN_VG_SIZE)
- if vgstatus:
- feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
+
+ # full package version
+ if constants.RELEASE_VERSION != remote_version[1]:
+ feedback_fn(" - WARNING: software version mismatch: master %s,"
+ " node %s %s" %
+ (constants.RELEASE_VERSION, node, remote_version[1]))
+
+ # checks vg existence and size > 20G
+ if vg_name is not None:
+ vglist = node_result.get(constants.NV_VGLIST, None)
+ if not vglist:
+ feedback_fn(" - ERROR: unable to check volume groups on node %s." %
+ (node,))
bad = True
+ else:
+ vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
+ constants.MIN_VG_SIZE)
+ if vgstatus:
+ feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
+ bad = True
# checks config file checksum
else:
# not candidate and this is not a must-have file
bad = True
- feedback_fn(" - ERROR: non master-candidate has old/wrong file"
- " '%s'" % file_name)
+ feedback_fn(" - ERROR: file '%s' should not exist on non master"
+ " candidates (and the file is outdated)" % file_name)
else:
# all good, except non-master/non-must have combination
if not node_is_mc and not must_have_file:
if hv_result is not None:
feedback_fn(" - ERROR: hypervisor %s verify failure: '%s'" %
(hv_name, hv_result))
+
+ # check used drbd list
+ if vg_name is not None:
+ used_minors = node_result.get(constants.NV_DRBDLIST, [])
+ if not isinstance(used_minors, (tuple, list)):
+ feedback_fn(" - ERROR: cannot parse drbd status file: %s" %
+ str(used_minors))
+ else:
+ for minor, (iname, must_exist) in drbd_map.items():
+ if minor not in used_minors and must_exist:
+ feedback_fn(" - ERROR: drbd minor %d of instance %s is"
+ " not active" % (minor, iname))
+ bad = True
+ for minor in used_minors:
+ if minor not in drbd_map:
+ feedback_fn(" - ERROR: unallocated drbd minor %d is in use" %
+ minor)
+ bad = True
+
return bad
def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
(volume, node))
bad = True
- if not instanceconfig.status == 'down':
+ if instanceconfig.admin_up:
if ((node_current not in node_instance or
not instance in node_instance[node_current]) and
node_current not in n_offline):
if bep[constants.BE_AUTO_BALANCE]:
needed_mem += bep[constants.BE_MEMORY]
if nodeinfo['mfree'] < needed_mem:
- feedback_fn(" - ERROR: not enough memory on node %s to accomodate"
+ feedback_fn(" - ERROR: not enough memory on node %s to accommodate"
" failovers should node %s fail" % (node, prinode))
bad = True
return bad
def BuildHooksEnv(self):
"""Build hooks env.
- Cluster-Verify hooks just rone in the post phase and their failure makes
+ Cluster-Verify hooks just ran in the post phase and their failure makes
the output be logged in the verify output and the verification to fail.
"""
all_nodes = self.cfg.GetNodeList()
- # TODO: populate the environment with useful information for verify hooks
- env = {}
+ env = {
+ "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
+ }
+ for node in self.cfg.GetAllNodesInfo().values():
+ env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
+
return env, [], all_nodes
def Exec(self, feedback_fn):
nodelist = utils.NiceSort(self.cfg.GetNodeList())
nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
instancelist = utils.NiceSort(self.cfg.GetInstanceList())
+ instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
+ for iname in instancelist)
i_non_redundant = [] # Non redundant instances
i_non_a_balanced = [] # Non auto-balanced instances
n_offline = [] # List of offline nodes
+ n_drained = [] # List of nodes being drained
node_volume = {}
node_instance = {}
node_info = {}
constants.NV_NODENETTEST: [(node.name, node.primary_ip,
node.secondary_ip) for node in nodeinfo
if not node.offline],
- constants.NV_LVLIST: vg_name,
constants.NV_INSTANCELIST: hypervisors,
- constants.NV_VGLIST: None,
constants.NV_VERSION: None,
constants.NV_HVINFO: self.cfg.GetHypervisorType(),
}
+ if vg_name is not None:
+ node_verify_param[constants.NV_VGLIST] = None
+ node_verify_param[constants.NV_LVLIST] = vg_name
+ node_verify_param[constants.NV_DRBDLIST] = None
all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
self.cfg.GetClusterName())
cluster = self.cfg.GetClusterInfo()
master_node = self.cfg.GetMasterNode()
+ all_drbd_map = self.cfg.ComputeDRBDMap()
+
for node_i in nodeinfo:
node = node_i.name
- nresult = all_nvinfo[node].data
if node_i.offline:
feedback_fn("* Skipping offline node %s" % (node,))
ntype = "master"
elif node_i.master_candidate:
ntype = "master candidate"
+ elif node_i.drained:
+ ntype = "drained"
+ n_drained.append(node)
else:
ntype = "regular"
feedback_fn("* Verifying node %s (%s)" % (node, ntype))
- if all_nvinfo[node].failed or not isinstance(nresult, dict):
- feedback_fn(" - ERROR: connection to %s failed" % (node,))
+ msg = all_nvinfo[node].fail_msg
+ if msg:
+ feedback_fn(" - ERROR: while contacting node %s: %s" % (node, msg))
bad = True
continue
+ nresult = all_nvinfo[node].payload
+ node_drbd = {}
+ for minor, instance in all_drbd_map[node].items():
+ if instance not in instanceinfo:
+ feedback_fn(" - ERROR: ghost instance '%s' in temporary DRBD map" %
+ instance)
+ # ghost instance should not be running, but otherwise we
+ # don't give double warnings (both ghost instance and
+ # unallocated minor in use)
+ node_drbd[minor] = (instance, False)
+ else:
+ instance = instanceinfo[instance]
+ node_drbd[minor] = (instance.name, instance.admin_up)
result = self._VerifyNode(node_i, file_names, local_checksums,
- nresult, feedback_fn, master_files)
+ nresult, feedback_fn, master_files,
+ node_drbd, vg_name)
bad = bad or result
lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
- if isinstance(lvdata, basestring):
+ if vg_name is None:
+ node_volume[node] = {}
+ elif isinstance(lvdata, basestring):
feedback_fn(" - ERROR: LVM problem on node %s: %s" %
- (node, lvdata.encode('string_escape')))
+ (node, utils.SafeEncode(lvdata)))
bad = True
node_volume[node] = {}
elif not isinstance(lvdata, dict):
try:
node_info[node] = {
"mfree": int(nodeinfo['memory_free']),
- "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
"pinst": [],
"sinst": [],
# dictionary holding all instances this node is secondary for,
# secondary.
"sinst-by-pnode": {},
}
- except ValueError:
- feedback_fn(" - ERROR: invalid value returned from node %s" % (node,))
+ # FIXME: devise a free space model for file based instances as well
+ if vg_name is not None:
+ if (constants.NV_VGLIST not in nresult or
+ vg_name not in nresult[constants.NV_VGLIST]):
+ feedback_fn(" - ERROR: node %s didn't return data for the"
+ " volume group '%s' - it is either missing or broken" %
+ (node, vg_name))
+ bad = True
+ continue
+ node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
+ except (ValueError, KeyError):
+ feedback_fn(" - ERROR: invalid nodeinfo value returned"
+ " from node %s" % (node,))
bad = True
continue
for instance in instancelist:
feedback_fn("* Verifying instance %s" % instance)
- inst_config = self.cfg.GetInstanceInfo(instance)
+ inst_config = instanceinfo[instance]
result = self._VerifyInstance(instance, inst_config, node_volume,
node_instance, feedback_fn, n_offline)
bad = bad or result
if n_offline:
feedback_fn(" - NOTICE: %d offline node(s) found." % len(n_offline))
+ if n_drained:
+ feedback_fn(" - NOTICE: %d drained node(s) found." % len(n_drained))
+
return not bad
def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
- """Analize the post-hooks' result
+ """Analyze the post-hooks' result
This method analyses the hook result, handles it, and sends some
nicely-formatted feedback back to the user.
for node_name in hooks_results:
show_node_header = True
res = hooks_results[node_name]
- if res.failed or res.data is False or not isinstance(res.data, list):
+ msg = res.fail_msg
+ if msg:
if res.offline:
# no need to warn or set fail return value
continue
- feedback_fn(" Communication failure in hooks execution")
+ feedback_fn(" Communication failure in hooks execution: %s" %
+ msg)
lu_result = 1
continue
- for script, hkr, output in res.data:
+ for script, hkr, output in res.payload:
if hkr == constants.HKR_FAIL:
# The node header is only shown once, if there are
# failing hooks on that node
locking.LEVEL_NODE: locking.ALL_SET,
locking.LEVEL_INSTANCE: locking.ALL_SET,
}
- self.share_locks = dict(((i, 1) for i in locking.LEVELS))
+ self.share_locks = dict.fromkeys(locking.LEVELS, 1)
def CheckPrereq(self):
"""Check prerequisites.
def Exec(self, feedback_fn):
"""Verify integrity of cluster disks.
+ @rtype: tuple of three items
+ @return: a tuple of (dict of node-to-node_error, list of instances
+ which need activate-disks, dict of instance: (node, volume) for
+ missing volumes
+
"""
- result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
+ result = res_nodes, res_instances, res_missing = {}, [], {}
vg_name = self.cfg.GetVGName()
nodes = utils.NiceSort(self.cfg.GetNodeList())
nv_dict = {}
for inst in instances:
inst_lvs = {}
- if (inst.status != "up" or
+ if (not inst.admin_up or
inst.disk_template not in constants.DTS_NET_MIRROR):
continue
inst.MapLVsByNode(inst_lvs)
if not nv_dict:
return result
- node_lvs = self.rpc.call_volume_list(nodes, vg_name)
+ node_lvs = self.rpc.call_lv_list(nodes, vg_name)
- to_act = set()
for node in nodes:
# node_volume
- lvs = node_lvs[node]
- if lvs.failed:
- if not lvs.offline:
- self.LogWarning("Connection to node %s failed: %s" %
- (node, lvs.data))
+ node_res = node_lvs[node]
+ if node_res.offline:
continue
- lvs = lvs.data
- if isinstance(lvs, basestring):
- logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
- res_nlvm[node] = lvs
- elif not isinstance(lvs, dict):
- logging.warning("Connection to node %s failed or invalid data"
- " returned", node)
- res_nodes.append(node)
+ msg = node_res.fail_msg
+ if msg:
+ logging.warning("Error enumerating LVs on node %s: %s", node, msg)
+ res_nodes[node] = msg
continue
- for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
+ lvs = node_res.payload
+ for lv_name, (_, lv_inactive, lv_online) in lvs.items():
inst = nv_dict.pop((node, lv_name), None)
if (not lv_online and inst is not None
and inst.name not in res_instances):
# shutdown the master IP
master = self.cfg.GetMasterNode()
result = self.rpc.call_node_stop_master(master, False)
- if result.failed or not result.data:
- raise errors.OpExecError("Could not disable the master role")
+ result.Raise("Could not disable the master role")
try:
cluster = self.cfg.GetClusterInfo()
result = self.rpc.call_upload_file(node_list,
constants.SSH_KNOWN_HOSTS_FILE)
for to_node, to_result in result.iteritems():
- if to_result.failed or not to_result.data:
- logging.error("Copy of file %s to node %s failed",
- constants.SSH_KNOWN_HOSTS_FILE, to_node)
+ msg = to_result.fail_msg
+ if msg:
+ msg = ("Copy of file %s to node %s failed: %s" %
+ (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
+ self.proc.LogWarning(msg)
finally:
- result = self.rpc.call_node_start_master(master, False)
- if result.failed or not result.data:
+ result = self.rpc.call_node_start_master(master, False, False)
+ msg = result.fail_msg
+ if msg:
self.LogWarning("Could not re-enable the master role on"
- " the master, please restart manually.")
+ " the master, please restart manually: %s", msg)
def _RecursiveCheckIfLVMBased(disk):
@type disk: L{objects.Disk}
@param disk: the disk to check
- @rtype: booleean
+ @rtype: boolean
@return: boolean indicating whether a LD_LV dev_type was found or not
"""
_OP_REQP = []
REQ_BGL = False
- def CheckParameters(self):
+ def CheckArguments(self):
"""Check parameters
"""
if self.op.candidate_pool_size is not None:
try:
self.op.candidate_pool_size = int(self.op.candidate_pool_size)
- except ValueError, err:
+ except (ValueError, TypeError), err:
raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
str(err))
if self.op.candidate_pool_size < 1:
if the given volume group is valid.
"""
- # FIXME: This only works because there is only one parameter that can be
- # changed or removed.
if self.op.vg_name is not None and not self.op.vg_name:
instances = self.cfg.GetAllInstancesInfo().values()
for inst in instances:
if self.op.vg_name:
vglist = self.rpc.call_vg_list(node_list)
for node in node_list:
- if vglist[node].failed:
+ msg = vglist[node].fail_msg
+ if msg:
# ignoring down node
- self.LogWarning("Node %s unreachable/error, ignoring" % node)
+ self.LogWarning("Error while gathering data on node %s"
+ " (ignoring node): %s", node, msg)
continue
- vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
+ vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
self.op.vg_name,
constants.MIN_VG_SIZE)
if vgstatus:
(node, vgstatus))
self.cluster = cluster = self.cfg.GetClusterInfo()
- # validate beparams changes
+ # validate params changes
if self.op.beparams:
- utils.CheckBEParams(self.op.beparams)
- self.new_beparams = cluster.FillDict(
- cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
+ utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
+ self.new_beparams = objects.FillDict(
+ cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
+
+ if self.op.nicparams:
+ utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
+ self.new_nicparams = objects.FillDict(
+ cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
+ objects.NIC.CheckParameterSyntax(self.new_nicparams)
# hypervisor list/parameters
- self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
+ self.new_hvparams = objects.FillDict(cluster.hvparams, {})
if self.op.hvparams:
if not isinstance(self.op.hvparams, dict):
raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
if self.op.enabled_hypervisors is not None:
self.hv_list = self.op.enabled_hypervisors
+ if not self.hv_list:
+ raise errors.OpPrereqError("Enabled hypervisors list must contain at"
+ " least one member")
+ invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
+ if invalid_hvs:
+ raise errors.OpPrereqError("Enabled hypervisors contains invalid"
+ " entries: %s" % invalid_hvs)
else:
self.hv_list = cluster.enabled_hypervisors
hv_name in self.op.enabled_hypervisors)):
# either this is a new hypervisor, or its parameters have changed
hv_class = hypervisor.GetHypervisor(hv_name)
+ utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
hv_class.CheckParameterSyntax(hv_params)
_CheckHVParams(self, node_list, hv_name, hv_params)
"""
if self.op.vg_name is not None:
- if self.op.vg_name != self.cfg.GetVGName():
- self.cfg.SetVGName(self.op.vg_name)
+ new_volume = self.op.vg_name
+ if not new_volume:
+ new_volume = None
+ if new_volume != self.cfg.GetVGName():
+ self.cfg.SetVGName(new_volume)
else:
feedback_fn("Cluster LVM configuration already in desired"
" state, not changing")
if self.op.enabled_hypervisors is not None:
self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
if self.op.beparams:
- self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
+ self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
+ if self.op.nicparams:
+ self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
+
if self.op.candidate_pool_size is not None:
self.cluster.candidate_pool_size = self.op.candidate_pool_size
+ # we need to update the pool size here, otherwise the save will fail
+ _AdjustCandidatePool(self)
self.cfg.Update(self.cluster)
- # we want to update nodes after the cluster so that if any errors
- # happen, we have recorded and saved the cluster info
- if self.op.candidate_pool_size is not None:
- _AdjustCandidatePool(self)
+
+def _RedistributeAncillaryFiles(lu, additional_nodes=None):
+ """Distribute additional files which are part of the cluster configuration.
+
+ ConfigWriter takes care of distributing the config and ssconf files, but
+ there are more files which should be distributed to all nodes. This function
+ makes sure those are copied.
+
+ @param lu: calling logical unit
+ @param additional_nodes: list of nodes not in the config to distribute to
+
+ """
+ # 1. Gather target nodes
+ myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
+ dist_nodes = lu.cfg.GetNodeList()
+ if additional_nodes is not None:
+ dist_nodes.extend(additional_nodes)
+ if myself.name in dist_nodes:
+ dist_nodes.remove(myself.name)
+ # 2. Gather files to distribute
+ dist_files = set([constants.ETC_HOSTS,
+ constants.SSH_KNOWN_HOSTS_FILE,
+ constants.RAPI_CERT_FILE,
+ constants.RAPI_USERS_FILE,
+ constants.HMAC_CLUSTER_KEY,
+ ])
+
+ enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
+ for hv_name in enabled_hypervisors:
+ hv_class = hypervisor.GetHypervisor(hv_name)
+ dist_files.update(hv_class.GetAncillaryFiles())
+
+ # 3. Perform the files upload
+ for fname in dist_files:
+ if os.path.exists(fname):
+ result = lu.rpc.call_upload_file(dist_nodes, fname)
+ for to_node, to_result in result.items():
+ msg = to_result.fail_msg
+ if msg:
+ msg = ("Copy of file %s to node %s failed: %s" %
+ (fname, to_node, msg))
+ lu.proc.LogWarning(msg)
class LURedistributeConfig(NoHooksLU):
"""
self.cfg.Update(self.cfg.GetClusterInfo())
+ _RedistributeAncillaryFiles(self)
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
lu.cfg.SetDiskID(dev, node)
retries = 0
+ degr_retries = 10 # in seconds, as we sleep 1 second each time
while True:
max_time = 0
done = True
cumul_degraded = False
rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
- if rstats.failed or not rstats.data:
- lu.LogWarning("Can't get any data from node %s", node)
+ msg = rstats.fail_msg
+ if msg:
+ lu.LogWarning("Can't get any data from node %s: %s", node, msg)
retries += 1
if retries >= 10:
raise errors.RemoteError("Can't contact node %s for mirror data,"
" aborting." % node)
time.sleep(6)
continue
- rstats = rstats.data
+ rstats = rstats.payload
retries = 0
for i, mstat in enumerate(rstats):
if mstat is None:
rem_time = "no time estimate"
lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
(instance.disks[i].iv_name, perc_done, rem_time))
+
+ # if we're done but degraded, let's do a few small retries, to
+ # make sure we see a stable and not transient situation; therefore
+ # we force restart of the loop
+ if (done or oneshot) and cumul_degraded and degr_retries > 0:
+ logging.info("Degraded disks found, %d retries left", degr_retries)
+ degr_retries -= 1
+ time.sleep(1)
+ continue
+
if done or oneshot:
break
result = True
if on_primary or dev.AssembleOnSecondary():
rstats = lu.rpc.call_blockdev_find(node, dev)
- if rstats.failed or not rstats.data:
- logging.warning("Node %s: disk degraded, not found or node down", node)
+ msg = rstats.fail_msg
+ if msg:
+ lu.LogWarning("Can't find disk on node %s: %s", node, msg)
+ result = False
+ elif not rstats.payload:
+ lu.LogWarning("Can't find disk on node %s", node)
result = False
else:
- result = result and (not rstats.data[idx])
+ result = result and (not rstats.payload[idx])
if dev.children:
for child in dev.children:
result = result and _CheckDiskConsistency(lu, child, node, on_primary)
selected=self.op.output_fields)
# Lock all nodes, in shared mode
+ # Temporary removal of locks, should be reverted later
+ # TODO: reintroduce locks when they are lighter-weight
self.needed_locks = {}
- self.share_locks[locking.LEVEL_NODE] = 1
- self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
+ #self.share_locks[locking.LEVEL_NODE] = 1
+ #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
def CheckPrereq(self):
"""Check prerequisites.
@param rlist: a map with node names as keys and OS objects as values
@rtype: dict
- @returns: a dictionary with osnames as keys and as value another map, with
- nodes as keys and list of OS objects as values, eg::
+ @return: a dictionary with osnames as keys and as value another map, with
+ nodes as keys and tuples of (path, status, diagnose) as values, eg::
- {"debian-etch": {"node1": [<object>,...],
- "node2": [<object>,]}
+ {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
+ (/srv/..., False, "invalid api")],
+ "node2": [(/srv/..., True, "")]}
}
"""
all_os = {}
- for node_name, nr in rlist.iteritems():
- if nr.failed or not nr.data:
+ # we build here the list of nodes that didn't fail the RPC (at RPC
+ # level), so that nodes with a non-responding node daemon don't
+ # make all OSes invalid
+ good_nodes = [node_name for node_name in rlist
+ if not rlist[node_name].fail_msg]
+ for node_name, nr in rlist.items():
+ if nr.fail_msg or not nr.payload:
continue
- for os_obj in nr.data:
- if os_obj.name not in all_os:
+ for name, path, status, diagnose in nr.payload:
+ if name not in all_os:
# build a list of nodes for this os containing empty lists
# for each node in node_list
- all_os[os_obj.name] = {}
- for nname in node_list:
- all_os[os_obj.name][nname] = []
- all_os[os_obj.name][node_name].append(os_obj)
+ all_os[name] = {}
+ for nname in good_nodes:
+ all_os[name][nname] = []
+ all_os[name][node_name].append((path, status, diagnose))
return all_os
def Exec(self, feedback_fn):
"""Compute the list of OSes.
"""
- node_list = self.acquired_locks[locking.LEVEL_NODE]
- valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
- if node in node_list]
+ valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
node_data = self.rpc.call_os_diagnose(valid_nodes)
- if node_data == False:
- raise errors.OpExecError("Can't gather the list of OSes")
pol = self._DiagnoseByOS(valid_nodes, node_data)
output = []
- for os_name, os_data in pol.iteritems():
+ for os_name, os_data in pol.items():
row = []
for field in self.op.output_fields:
if field == "name":
val = os_name
elif field == "valid":
- val = utils.all([osl and osl[0] for osl in os_data.values()])
+ val = utils.all([osl and osl[0][1] for osl in os_data.values()])
elif field == "node_status":
+ # this is just a copy of the dict
val = {}
- for node_name, nos_list in os_data.iteritems():
- val[node_name] = [(v.status, v.path) for v in nos_list]
+ for node_name, nos_list in os_data.items():
+ val[node_name] = nos_list
else:
raise errors.ParameterError(field)
row.append(val)
- it does not have primary or secondary instances
- it's not the master
- Any errors are signalled by raising errors.OpPrereqError.
+ Any errors are signaled by raising errors.OpPrereqError.
"""
node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
self.context.RemoveNode(node.name)
- self.rpc.call_node_leave_cluster(node.name)
+ result = self.rpc.call_node_leave_cluster(node.name)
+ msg = result.fail_msg
+ if msg:
+ self.LogWarning("Errors encountered on the remote node while leaving"
+ " the cluster: %s", msg)
# Promote nodes to master candidate as needed
_AdjustCandidatePool(self)
"""Logical unit for querying nodes.
"""
- _OP_REQP = ["output_fields", "names"]
+ _OP_REQP = ["output_fields", "names", "use_locking"]
REQ_BGL = False
_FIELDS_DYNAMIC = utils.FieldSet(
"dtotal", "dfree",
"mtotal", "mnode", "mfree",
"bootid",
- "ctotal",
+ "ctotal", "cnodes", "csockets",
)
_FIELDS_STATIC = utils.FieldSet(
"master_candidate",
"master",
"offline",
+ "drained",
+ "role",
)
def ExpandNames(self):
else:
self.wanted = locking.ALL_SET
- self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
+ self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
+ self.do_locking = self.do_node_query and self.op.use_locking
if self.do_locking:
# if we don't request only static fields, we need to lock the nodes
self.needed_locks[locking.LEVEL_NODE] = self.wanted
# begin data gathering
- if self.do_locking:
+ if self.do_node_query:
live_data = {}
node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
self.cfg.GetHypervisorType())
for name in nodenames:
nodeinfo = node_data[name]
- if not nodeinfo.failed and nodeinfo.data:
- nodeinfo = nodeinfo.data
+ if not nodeinfo.fail_msg and nodeinfo.payload:
+ nodeinfo = nodeinfo.payload
fn = utils.TryConvert
live_data[name] = {
"mtotal": fn(int, nodeinfo.get('memory_total', None)),
"dfree": fn(int, nodeinfo.get('vg_free', None)),
"ctotal": fn(int, nodeinfo.get('cpu_total', None)),
"bootid": nodeinfo.get('bootid', None),
+ "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
+ "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
}
else:
live_data[name] = {}
val = node.name == master_node
elif field == "offline":
val = node.offline
+ elif field == "drained":
+ val = node.drained
elif self._FIELDS_DYNAMIC.Matches(field):
val = live_data[node.name].get(field, None)
+ elif field == "role":
+ if node.name == master_node:
+ val = "M"
+ elif node.master_candidate:
+ val = "C"
+ elif node.drained:
+ val = "D"
+ elif node.offline:
+ val = "O"
+ else:
+ val = "R"
else:
raise errors.ParameterError(field)
node_output.append(val)
output = []
for node in nodenames:
- if node not in volumes or volumes[node].failed or not volumes[node].data:
+ nresult = volumes[node]
+ if nresult.offline:
+ continue
+ msg = nresult.fail_msg
+ if msg:
+ self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
continue
- node_vols = volumes[node].data[:]
+ node_vols = nresult.payload[:]
node_vols.sort(key=lambda vol: vol['dev'])
for vol in node_vols:
- it is resolvable
- its parameters (single/dual homed) matches the cluster
- Any errors are signalled by raising errors.OpPrereqError.
+ Any errors are signaled by raising errors.OpPrereqError.
"""
node_name = self.op.node_name
raise errors.OpPrereqError("The master has a private ip but the"
" new node doesn't have one")
- # checks reachablity
+ # checks reachability
if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
raise errors.OpPrereqError("Node not reachable by ping")
" based ping to noded port")
cp_size = self.cfg.GetClusterInfo().candidate_pool_size
- mc_now, _ = self.cfg.GetMasterCandidateStats()
- master_candidate = mc_now < cp_size
+ if self.op.readd:
+ exceptions = [node]
+ else:
+ exceptions = []
+ mc_now, mc_max = self.cfg.GetMasterCandidateStats(exceptions)
+ # the new node will increase mc_max with one, so:
+ mc_max = min(mc_max + 1, cp_size)
+ self.master_candidate = mc_now < mc_max
- self.new_node = objects.Node(name=node,
- primary_ip=primary_ip,
- secondary_ip=secondary_ip,
- master_candidate=master_candidate,
- offline=False)
+ if self.op.readd:
+ self.new_node = self.cfg.GetNodeInfo(node)
+ assert self.new_node is not None, "Can't retrieve locked node %s" % node
+ else:
+ self.new_node = objects.Node(name=node,
+ primary_ip=primary_ip,
+ secondary_ip=secondary_ip,
+ master_candidate=self.master_candidate,
+ offline=False, drained=False)
def Exec(self, feedback_fn):
"""Adds the new node to the cluster.
new_node = self.new_node
node = new_node.name
+ # for re-adds, reset the offline/drained/master-candidate flags;
+ # we need to reset here, otherwise offline would prevent RPC calls
+ # later in the procedure; this also means that if the re-add
+ # fails, we are left with a non-offlined, broken node
+ if self.op.readd:
+ new_node.drained = new_node.offline = False
+ self.LogInfo("Readding a node, the offline/drained flags were reset")
+ # if we demote the node, we do cleanup later in the procedure
+ new_node.master_candidate = self.master_candidate
+
+ # notify the user about any possible mc promotion
+ if new_node.master_candidate:
+ self.LogInfo("Node will be a master candidate")
+
# check connectivity
result = self.rpc.call_version([node])[node]
- result.Raise()
- if result.data:
- if constants.PROTOCOL_VERSION == result.data:
- logging.info("Communication to node %s fine, sw version %s match",
- node, result.data)
- else:
- raise errors.OpExecError("Version mismatch master version %s,"
- " node version %s" %
- (constants.PROTOCOL_VERSION, result.data))
+ result.Raise("Can't get version information from node %s" % node)
+ if constants.PROTOCOL_VERSION == result.payload:
+ logging.info("Communication to node %s fine, sw version %s match",
+ node, result.payload)
else:
- raise errors.OpExecError("Cannot get version from the new node")
+ raise errors.OpExecError("Version mismatch master version %s,"
+ " node version %s" %
+ (constants.PROTOCOL_VERSION, result.payload))
# setup ssh on node
logging.info("Copy ssh key to node %s", node)
result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
keyarray[2],
keyarray[3], keyarray[4], keyarray[5])
-
- if result.failed or not result.data:
- raise errors.OpExecError("Cannot transfer ssh keys to the new node")
+ result.Raise("Cannot transfer ssh keys to the new node")
# Add node to our /etc/hosts, and add key to known_hosts
- utils.AddHostToEtcHosts(new_node.name)
+ if self.cfg.GetClusterInfo().modify_etc_hosts:
+ utils.AddHostToEtcHosts(new_node.name)
if new_node.secondary_ip != new_node.primary_ip:
result = self.rpc.call_node_has_ip_address(new_node.name,
new_node.secondary_ip)
- if result.failed or not result.data:
+ result.Raise("Failure checking secondary ip on node %s" % new_node.name,
+ prereq=True)
+ if not result.payload:
raise errors.OpExecError("Node claims it doesn't have the secondary ip"
" you gave (%s). Please fix and re-run this"
" command." % new_node.secondary_ip)
result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
self.cfg.GetClusterName())
for verifier in node_verify_list:
- if result[verifier].failed or not result[verifier].data:
- raise errors.OpExecError("Cannot communicate with %s's node daemon"
- " for remote verification" % verifier)
- if result[verifier].data['nodelist']:
- for failed in result[verifier].data['nodelist']:
+ result[verifier].Raise("Cannot communicate with node %s" % verifier)
+ nl_payload = result[verifier].payload['nodelist']
+ if nl_payload:
+ for failed in nl_payload:
feedback_fn("ssh/hostname verification failed %s -> %s" %
- (verifier, result[verifier]['nodelist'][failed]))
+ (verifier, nl_payload[failed]))
raise errors.OpExecError("ssh/hostname verification failed.")
- # Distribute updated /etc/hosts and known_hosts to all nodes,
- # including the node just added
- myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
- dist_nodes = self.cfg.GetNodeList()
- if not self.op.readd:
- dist_nodes.append(node)
- if myself.name in dist_nodes:
- dist_nodes.remove(myself.name)
-
- logging.debug("Copying hosts and known_hosts to all nodes")
- for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
- result = self.rpc.call_upload_file(dist_nodes, fname)
- for to_node, to_result in result.iteritems():
- if to_result.failed or not to_result.data:
- logging.error("Copy of file %s to node %s failed", fname, to_node)
-
- to_copy = []
- if constants.HT_XEN_HVM in self.cfg.GetClusterInfo().enabled_hypervisors:
- to_copy.append(constants.VNC_PASSWORD_FILE)
- for fname in to_copy:
- result = self.rpc.call_upload_file([node], fname)
- if result[node].failed or not result[node]:
- logging.error("Could not copy file %s to node %s", fname, node)
-
if self.op.readd:
+ _RedistributeAncillaryFiles(self)
self.context.ReaddNode(new_node)
+ # make sure we redistribute the config
+ self.cfg.Update(new_node)
+ # and make sure the new node will not have old files around
+ if not new_node.master_candidate:
+ result = self.rpc.call_node_demote_from_mc(new_node.name)
+ msg = result.RemoteFailMsg()
+ if msg:
+ self.LogWarning("Node failed to demote itself from master"
+ " candidate status: %s" % msg)
else:
+ _RedistributeAncillaryFiles(self, additional_nodes=[node])
self.context.AddNode(new_node)
self.op.node_name = node_name
_CheckBooleanOpField(self.op, 'master_candidate')
_CheckBooleanOpField(self.op, 'offline')
- if self.op.master_candidate is None and self.op.offline is None:
+ _CheckBooleanOpField(self.op, 'drained')
+ all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
+ if all_mods.count(None) == 3:
raise errors.OpPrereqError("Please pass at least one modification")
- if self.op.offline == True and self.op.master_candidate == True:
- raise errors.OpPrereqError("Can't set the node into offline and"
- " master_candidate at the same time")
+ if all_mods.count(True) > 1:
+ raise errors.OpPrereqError("Can't set the node into more than one"
+ " state at the same time")
def ExpandNames(self):
self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
"OP_TARGET": self.op.node_name,
"MASTER_CANDIDATE": str(self.op.master_candidate),
"OFFLINE": str(self.op.offline),
+ "DRAINED": str(self.op.drained),
}
nl = [self.cfg.GetMasterNode(),
self.op.node_name]
"""
node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
- if ((self.op.master_candidate == False or self.op.offline == True)
- and node.master_candidate):
+ if ((self.op.master_candidate == False or self.op.offline == True or
+ self.op.drained == True) and node.master_candidate):
# we will demote the node from master_candidate
if self.op.node_name == self.cfg.GetMasterNode():
raise errors.OpPrereqError("The master node has to be a"
- " master candidate and online")
+ " master candidate, online and not drained")
cp_size = self.cfg.GetClusterInfo().candidate_pool_size
num_candidates, _ = self.cfg.GetMasterCandidateStats()
if num_candidates <= cp_size:
else:
raise errors.OpPrereqError(msg)
- if (self.op.master_candidate == True and node.offline and
- not self.op.offline == False):
- raise errors.OpPrereqError("Can't set an offline node to"
- " master_candidate")
+ if (self.op.master_candidate == True and
+ ((node.offline and not self.op.offline == False) or
+ (node.drained and not self.op.drained == False))):
+ raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
+ " to master_candidate" % node.name)
return
node = self.node
result = []
+ changed_mc = False
if self.op.offline is not None:
node.offline = self.op.offline
result.append(("offline", str(self.op.offline)))
- if self.op.offline == True and node.master_candidate:
- node.master_candidate = False
- result.append(("master_candidate", "auto-demotion due to offline"))
+ if self.op.offline == True:
+ if node.master_candidate:
+ node.master_candidate = False
+ changed_mc = True
+ result.append(("master_candidate", "auto-demotion due to offline"))
+ if node.drained:
+ node.drained = False
+ result.append(("drained", "clear drained status due to offline"))
if self.op.master_candidate is not None:
node.master_candidate = self.op.master_candidate
+ changed_mc = True
result.append(("master_candidate", str(self.op.master_candidate)))
if self.op.master_candidate == False:
rrc = self.rpc.call_node_demote_from_mc(node.name)
- if (rrc.failed or not isinstance(rrc.data, (tuple, list))
- or len(rrc.data) != 2):
- self.LogWarning("Node rpc error: %s" % rrc.error)
- elif not rrc.data[0]:
- self.LogWarning("Node failed to demote itself: %s" % rrc.data[1])
+ msg = rrc.fail_msg
+ if msg:
+ self.LogWarning("Node failed to demote itself: %s" % msg)
+
+ if self.op.drained is not None:
+ node.drained = self.op.drained
+ result.append(("drained", str(self.op.drained)))
+ if self.op.drained == True:
+ if node.master_candidate:
+ node.master_candidate = False
+ changed_mc = True
+ result.append(("master_candidate", "auto-demotion due to drain"))
+ rrc = self.rpc.call_node_demote_from_mc(node.name)
+ msg = rrc.RemoteFailMsg()
+ if msg:
+ self.LogWarning("Node failed to demote itself: %s" % msg)
+ if node.offline:
+ node.offline = False
+ result.append(("offline", "clear offline status due to drain"))
# this will trigger configuration file update, if needed
self.cfg.Update(node)
# this will trigger job queue propagation or cleanup
- if self.op.node_name != self.cfg.GetMasterNode():
+ if changed_mc:
self.context.ReaddNode(node)
return result
+class LUPowercycleNode(NoHooksLU):
+ """Powercycles a node.
+
+ """
+ _OP_REQP = ["node_name", "force"]
+ REQ_BGL = False
+
+ def CheckArguments(self):
+ node_name = self.cfg.ExpandNodeName(self.op.node_name)
+ if node_name is None:
+ raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
+ self.op.node_name = node_name
+ if node_name == self.cfg.GetMasterNode() and not self.op.force:
+ raise errors.OpPrereqError("The node is the master and the force"
+ " parameter was not set")
+
+ def ExpandNames(self):
+ """Locking for PowercycleNode.
+
+ This is a last-resource option and shouldn't block on other
+ jobs. Therefore, we grab no locks.
+
+ """
+ self.needed_locks = {}
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This LU has no prereqs.
+
+ """
+ pass
+
+ def Exec(self, feedback_fn):
+ """Reboots a node.
+
+ """
+ result = self.rpc.call_node_powercycle(self.op.node_name,
+ self.cfg.GetHypervisorType())
+ result.Raise("Failed to schedule the reboot")
+ return result.payload
+
+
class LUQueryClusterInfo(NoHooksLU):
"""Query cluster configuration.
"software_version": constants.RELEASE_VERSION,
"protocol_version": constants.PROTOCOL_VERSION,
"config_version": constants.CONFIG_VERSION,
- "os_api_version": constants.OS_API_VERSION,
+ "os_api_version": max(constants.OS_API_VERSIONS),
"export_version": constants.EXPORT_VERSION,
"architecture": (platform.architecture()[0], platform.machine()),
"name": cluster.cluster_name,
"master": cluster.master_node,
- "default_hypervisor": cluster.default_hypervisor,
+ "default_hypervisor": cluster.enabled_hypervisors[0],
"enabled_hypervisors": cluster.enabled_hypervisors,
- "hvparams": cluster.hvparams,
+ "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
+ for hypervisor_name in cluster.enabled_hypervisors]),
"beparams": cluster.beparams,
+ "nicparams": cluster.nicparams,
"candidate_pool_size": cluster.candidate_pool_size,
+ "master_netdev": cluster.master_netdev,
+ "volume_group_name": cluster.volume_group_name,
+ "file_storage_dir": cluster.file_storage_dir,
}
return result
for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
lu.cfg.SetDiskID(node_disk, node)
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
- if result.failed or not result:
+ msg = result.fail_msg
+ if msg:
lu.proc.LogWarning("Could not prepare block device %s on node %s"
- " (is_primary=False, pass=1)",
- inst_disk.iv_name, node)
+ " (is_primary=False, pass=1): %s",
+ inst_disk.iv_name, node, msg)
if not ignore_secondaries:
disks_ok = False
continue
lu.cfg.SetDiskID(node_disk, node)
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
- if result.failed or not result:
+ msg = result.fail_msg
+ if msg:
lu.proc.LogWarning("Could not prepare block device %s on node %s"
- " (is_primary=True, pass=2)",
- inst_disk.iv_name, node)
+ " (is_primary=True, pass=2): %s",
+ inst_disk.iv_name, node, msg)
disks_ok = False
- device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
+ device_info.append((instance.primary_node, inst_disk.iv_name,
+ result.payload))
# leave the disks configured for the primary node
# this is a workaround that would be fixed better by
"""Start the disks of an instance.
"""
- disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
+ disks_ok, _ = _AssembleInstanceDisks(lu, instance,
ignore_secondaries=force)
if not disks_ok:
_ShutdownInstanceDisks(lu, instance)
_ShutdownInstanceDisks.
"""
- ins_l = lu.rpc.call_instance_list([instance.primary_node],
- [instance.hypervisor])
- ins_l = ins_l[instance.primary_node]
- if ins_l.failed or not isinstance(ins_l.data, list):
- raise errors.OpExecError("Can't contact node '%s'" %
- instance.primary_node)
-
- if instance.name in ins_l.data:
+ pnode = instance.primary_node
+ ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
+ ins_l.Raise("Can't contact node %s" % pnode)
+
+ if instance.name in ins_l.payload:
raise errors.OpExecError("Instance is running, can't shutdown"
" block devices.")
ignored.
"""
- result = True
+ all_result = True
for disk in instance.disks:
for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
lu.cfg.SetDiskID(top_disk, node)
result = lu.rpc.call_blockdev_shutdown(node, top_disk)
- if result.failed or not result.data:
- logging.error("Could not shutdown block device %s on node %s",
- disk.iv_name, node)
+ msg = result.fail_msg
+ if msg:
+ lu.LogWarning("Could not shutdown block device %s on node %s: %s",
+ disk.iv_name, node, msg)
if not ignore_primary or node != instance.primary_node:
- result = False
- return result
+ all_result = False
+ return all_result
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
"""
nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
- nodeinfo[node].Raise()
- free_mem = nodeinfo[node].data.get('memory_free')
+ nodeinfo[node].Raise("Can't get data from node %s" % node, prereq=True)
+ free_mem = nodeinfo[node].payload.get('memory_free', None)
if not isinstance(free_mem, int):
raise errors.OpPrereqError("Can't compute free memory on node %s, result"
- " was '%s'" % (node, free_mem))
+ " was '%s'" % (node, free_mem))
if requested > free_mem:
raise errors.OpPrereqError("Not enough memory on node %s for %s:"
- " needed %s MiB, available %s MiB" %
- (node, reason, requested, free_mem))
+ " needed %s MiB, available %s MiB" %
+ (node, reason, requested, free_mem))
class LUStartupInstance(LogicalUnit):
assert self.instance is not None, \
"Cannot retrieve locked instance %s" % self.op.instance_name
+ # extra beparams
+ self.beparams = getattr(self.op, "beparams", {})
+ if self.beparams:
+ if not isinstance(self.beparams, dict):
+ raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
+ " dict" % (type(self.beparams), ))
+ # fill the beparams dict
+ utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
+ self.op.beparams = self.beparams
+
+ # extra hvparams
+ self.hvparams = getattr(self.op, "hvparams", {})
+ if self.hvparams:
+ if not isinstance(self.hvparams, dict):
+ raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
+ " dict" % (type(self.hvparams), ))
+
+ # check hypervisor parameter syntax (locally)
+ cluster = self.cfg.GetClusterInfo()
+ utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
+ filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
+ instance.hvparams)
+ filled_hvp.update(self.hvparams)
+ hv_type = hypervisor.GetHypervisor(instance.hypervisor)
+ hv_type.CheckParameterSyntax(filled_hvp)
+ _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
+ self.op.hvparams = self.hvparams
+
_CheckNodeOnline(self, instance.primary_node)
bep = self.cfg.GetClusterInfo().FillBE(instance)
- # check bridges existance
+ # check bridges existence
_CheckInstanceBridgesExist(self, instance)
- _CheckNodeFreeMemory(self, instance.primary_node,
- "starting instance %s" % instance.name,
- bep[constants.BE_MEMORY], instance.hypervisor)
+ remote_info = self.rpc.call_instance_info(instance.primary_node,
+ instance.name,
+ instance.hypervisor)
+ remote_info.Raise("Error checking node %s" % instance.primary_node,
+ prereq=True)
+ if not remote_info.payload: # not running already
+ _CheckNodeFreeMemory(self, instance.primary_node,
+ "starting instance %s" % instance.name,
+ bep[constants.BE_MEMORY], instance.hypervisor)
def Exec(self, feedback_fn):
"""Start the instance.
"""
instance = self.instance
force = self.op.force
- extra_args = getattr(self.op, "extra_args", "")
self.cfg.MarkInstanceUp(instance.name)
_StartInstanceDisks(self, instance, force)
- result = self.rpc.call_instance_start(node_current, instance, extra_args)
- msg = result.RemoteFailMsg()
+ result = self.rpc.call_instance_start(node_current, instance,
+ self.hvparams, self.beparams)
+ msg = result.fail_msg
if msg:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Could not start instance: %s" % msg)
"""
env = {
"IGNORE_SECONDARIES": self.op.ignore_secondaries,
+ "REBOOT_TYPE": self.op.reboot_type,
}
env.update(_BuildInstanceHookEnvByObject(self, self.instance))
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
_CheckNodeOnline(self, instance.primary_node)
- # check bridges existance
+ # check bridges existence
_CheckInstanceBridgesExist(self, instance)
def Exec(self, feedback_fn):
instance = self.instance
ignore_secondaries = self.op.ignore_secondaries
reboot_type = self.op.reboot_type
- extra_args = getattr(self.op, "extra_args", "")
node_current = instance.primary_node
if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
constants.INSTANCE_REBOOT_HARD]:
+ for disk in instance.disks:
+ self.cfg.SetDiskID(disk, node_current)
result = self.rpc.call_instance_reboot(node_current, instance,
- reboot_type, extra_args)
- if result.failed or not result.data:
- raise errors.OpExecError("Could not reboot instance")
+ reboot_type)
+ result.Raise("Could not reboot instance")
else:
- if not self.rpc.call_instance_shutdown(node_current, instance):
- raise errors.OpExecError("could not shutdown instance for full reboot")
+ result = self.rpc.call_instance_shutdown(node_current, instance)
+ result.Raise("Could not shutdown instance for full reboot")
_ShutdownInstanceDisks(self, instance)
_StartInstanceDisks(self, instance, ignore_secondaries)
- result = self.rpc.call_instance_start(node_current, instance, extra_args)
- msg = result.RemoteFailMsg()
+ result = self.rpc.call_instance_start(node_current, instance, None, None)
+ msg = result.fail_msg
if msg:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Could not start instance for"
node_current = instance.primary_node
self.cfg.MarkInstanceDown(instance.name)
result = self.rpc.call_instance_shutdown(node_current, instance)
- if result.failed or not result.data:
- self.proc.LogWarning("Could not shutdown instance")
+ msg = result.fail_msg
+ if msg:
+ self.proc.LogWarning("Could not shutdown instance: %s" % msg)
_ShutdownInstanceDisks(self, instance)
if instance.disk_template == constants.DT_DISKLESS:
raise errors.OpPrereqError("Instance '%s' has no disks" %
self.op.instance_name)
- if instance.status != "down":
+ if instance.admin_up:
raise errors.OpPrereqError("Instance '%s' is marked to be up" %
self.op.instance_name)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- if remote_info.failed or remote_info.data:
+ remote_info.Raise("Error checking node %s" % instance.primary_node,
+ prereq=True)
+ if remote_info.payload:
raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
(self.op.instance_name,
instance.primary_node))
raise errors.OpPrereqError("Primary node '%s' is unknown" %
self.op.pnode)
result = self.rpc.call_os_get(pnode.name, self.op.os_type)
- result.Raise()
- if not isinstance(result.data, objects.OS):
- raise errors.OpPrereqError("OS '%s' not in supported OS list for"
- " primary node" % self.op.os_type)
+ result.Raise("OS '%s' not in supported OS list for primary node %s" %
+ (self.op.os_type, pnode.name), prereq=True)
self.instance = instance
_StartInstanceDisks(self, inst, None)
try:
feedback_fn("Running the instance OS create scripts...")
- result = self.rpc.call_instance_os_add(inst.primary_node, inst)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not install OS for instance %s"
- " on node %s: %s" %
- (inst.name, inst.primary_node, msg))
+ result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
+ result.Raise("Could not install OS for instance %s on node %s" %
+ (inst.name, inst.primary_node))
finally:
_ShutdownInstanceDisks(self, inst)
self.op.instance_name)
_CheckNodeOnline(self, instance.primary_node)
- if instance.status != "down":
+ if instance.admin_up:
raise errors.OpPrereqError("Instance '%s' is marked to be up" %
self.op.instance_name)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- remote_info.Raise()
- if remote_info.data:
+ remote_info.Raise("Error checking node %s" % instance.primary_node,
+ prereq=True)
+ if remote_info.payload:
raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
(self.op.instance_name,
instance.primary_node))
result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
old_file_storage_dir,
new_file_storage_dir)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Could not connect to node '%s' to rename"
- " directory '%s' to '%s' (but the instance"
- " has been renamed in Ganeti)" % (
- inst.primary_node, old_file_storage_dir,
- new_file_storage_dir))
-
- if not result.data[0]:
- raise errors.OpExecError("Could not rename directory '%s' to '%s'"
- " (but the instance has been renamed in"
- " Ganeti)" % (old_file_storage_dir,
- new_file_storage_dir))
+ result.Raise("Could not rename on node %s directory '%s' to '%s'"
+ " (but the instance has been renamed in Ganeti)" %
+ (inst.primary_node, old_file_storage_dir,
+ new_file_storage_dir))
_StartInstanceDisks(self, inst, None)
try:
result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
old_name)
- if result.failed or not result.data:
+ msg = result.fail_msg
+ if msg:
msg = ("Could not run OS rename script for instance %s on node %s"
- " (but the instance has been renamed in Ganeti)" %
- (inst.name, inst.primary_node))
+ " (but the instance has been renamed in Ganeti): %s" %
+ (inst.name, inst.primary_node, msg))
self.proc.LogWarning(msg)
finally:
_ShutdownInstanceDisks(self, inst)
instance.name, instance.primary_node)
result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
- if result.failed or not result.data:
+ msg = result.fail_msg
+ if msg:
if self.op.ignore_failures:
- feedback_fn("Warning: can't shutdown instance")
+ feedback_fn("Warning: can't shutdown instance: %s" % msg)
else:
- raise errors.OpExecError("Could not shutdown instance %s on node %s" %
- (instance.name, instance.primary_node))
+ raise errors.OpExecError("Could not shutdown instance %s on"
+ " node %s: %s" %
+ (instance.name, instance.primary_node, msg))
logging.info("Removing block devices for instance %s", instance.name)
"""Logical unit for querying instances.
"""
- _OP_REQP = ["output_fields", "names"]
+ _OP_REQP = ["output_fields", "names", "use_locking"]
REQ_BGL = False
_FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
- "admin_state", "admin_ram",
+ "admin_state",
"disk_template", "ip", "mac", "bridge",
+ "nic_mode", "nic_link",
"sda_size", "sdb_size", "vcpus", "tags",
"network_port", "beparams",
- "(disk).(size)/([0-9]+)",
- "(disk).(sizes)",
- "(nic).(mac|ip|bridge)/([0-9]+)",
- "(nic).(macs|ips|bridges)",
- "(disk|nic).(count)",
+ r"(disk)\.(size)/([0-9]+)",
+ r"(disk)\.(sizes)", "disk_usage",
+ r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
+ r"(nic)\.(bridge)/([0-9]+)",
+ r"(nic)\.(macs|ips|modes|links|bridges)",
+ r"(disk|nic)\.(count)",
"serial_no", "hypervisor", "hvparams",] +
["hv/%s" % name
for name in constants.HVS_PARAMETERS] +
else:
self.wanted = locking.ALL_SET
- self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
+ self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
+ self.do_locking = self.do_node_query and self.op.use_locking
if self.do_locking:
self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
self.needed_locks[locking.LEVEL_NODE] = []
"""
all_info = self.cfg.GetAllInstancesInfo()
- if self.do_locking:
- instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
- elif self.wanted != locking.ALL_SET:
- instance_names = self.wanted
- missing = set(instance_names).difference(all_info.keys())
- if missing:
- raise errors.OpExecError(
- "Some instances were removed before retrieving their data: %s"
- % missing)
+ if self.wanted == locking.ALL_SET:
+ # caller didn't specify instance names, so ordering is not important
+ if self.do_locking:
+ instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
+ else:
+ instance_names = all_info.keys()
+ instance_names = utils.NiceSort(instance_names)
else:
- instance_names = all_info.keys()
+ # caller did specify names, so we must keep the ordering
+ if self.do_locking:
+ tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
+ else:
+ tgt_set = all_info.keys()
+ missing = set(self.wanted).difference(tgt_set)
+ if missing:
+ raise errors.OpExecError("Some instances were removed before"
+ " retrieving their data: %s" % missing)
+ instance_names = self.wanted
- instance_names = utils.NiceSort(instance_names)
instance_list = [all_info[iname] for iname in instance_names]
# begin data gathering
bad_nodes = []
off_nodes = []
- if self.do_locking:
+ if self.do_node_query:
live_data = {}
node_data = self.rpc.call_all_instances_info(nodes, hv_list)
for name in nodes:
if result.offline:
# offline nodes will be in both lists
off_nodes.append(name)
- if result.failed:
+ if result.failed or result.fail_msg:
bad_nodes.append(name)
else:
- if result.data:
- live_data.update(result.data)
- # else no instance is alive
+ if result.payload:
+ live_data.update(result.payload)
+ # else no instance is alive
else:
live_data = dict([(name, {}) for name in instance_names])
HVPREFIX = "hv/"
BEPREFIX = "be/"
output = []
+ cluster = self.cfg.GetClusterInfo()
for instance in instance_list:
iout = []
- i_hv = self.cfg.GetClusterInfo().FillHV(instance)
- i_be = self.cfg.GetClusterInfo().FillBE(instance)
+ i_hv = cluster.FillHV(instance)
+ i_be = cluster.FillBE(instance)
+ i_nicp = [objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
+ nic.nicparams) for nic in instance.nics]
for field in self.op.output_fields:
st_match = self._FIELDS_STATIC.Matches(field)
if field == "name":
elif field == "snodes":
val = list(instance.secondary_nodes)
elif field == "admin_state":
- val = (instance.status != "down")
+ val = instance.admin_up
elif field == "oper_state":
if instance.primary_node in bad_nodes:
val = None
else:
running = bool(live_data.get(instance.name))
if running:
- if instance.status != "down":
+ if instance.admin_up:
val = "running"
else:
val = "ERROR_up"
else:
- if instance.status != "down":
+ if instance.admin_up:
val = "ERROR_down"
else:
val = "ADMIN_down"
val = live_data[instance.name].get("memory", "?")
else:
val = "-"
+ elif field == "vcpus":
+ val = i_be[constants.BE_VCPUS]
elif field == "disk_template":
val = instance.disk_template
elif field == "ip":
- val = instance.nics[0].ip
+ if instance.nics:
+ val = instance.nics[0].ip
+ else:
+ val = None
+ elif field == "nic_mode":
+ if instance.nics:
+ val = i_nicp[0][constants.NIC_MODE]
+ else:
+ val = None
+ elif field == "nic_link":
+ if instance.nics:
+ val = i_nicp[0][constants.NIC_LINK]
+ else:
+ val = None
elif field == "bridge":
- val = instance.nics[0].bridge
+ if (instance.nics and
+ i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
+ val = i_nicp[0][constants.NIC_LINK]
+ else:
+ val = None
elif field == "mac":
- val = instance.nics[0].mac
+ if instance.nics:
+ val = instance.nics[0].mac
+ else:
+ val = None
elif field == "sda_size" or field == "sdb_size":
idx = ord(field[2]) - ord('a')
try:
val = instance.FindDisk(idx).size
except errors.OpPrereqError:
val = None
+ elif field == "disk_usage": # total disk usage per node
+ disk_sizes = [{'size': disk.size} for disk in instance.disks]
+ val = _ComputeDiskSize(instance.disk_template, disk_sizes)
elif field == "tags":
val = list(instance.GetTags())
elif field == "serial_no":
val = [nic.mac for nic in instance.nics]
elif st_groups[1] == "ips":
val = [nic.ip for nic in instance.nics]
+ elif st_groups[1] == "modes":
+ val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
+ elif st_groups[1] == "links":
+ val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
elif st_groups[1] == "bridges":
- val = [nic.bridge for nic in instance.nics]
+ val = []
+ for nicp in i_nicp:
+ if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
+ val.append(nicp[constants.NIC_LINK])
+ else:
+ val.append(None)
else:
# index-based item
nic_idx = int(st_groups[2])
val = instance.nics[nic_idx].mac
elif st_groups[1] == "ip":
val = instance.nics[nic_idx].ip
+ elif st_groups[1] == "mode":
+ val = i_nicp[nic_idx][constants.NIC_MODE]
+ elif st_groups[1] == "link":
+ val = i_nicp[nic_idx][constants.NIC_LINK]
elif st_groups[1] == "bridge":
- val = instance.nics[nic_idx].bridge
+ nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
+ if nic_mode == constants.NIC_MODE_BRIDGED:
+ val = i_nicp[nic_idx][constants.NIC_LINK]
+ else:
+ val = None
else:
assert False, "Unhandled NIC parameter"
else:
- assert False, "Unhandled variable parameter"
+ assert False, ("Declared but unhandled variable parameter '%s'" %
+ field)
else:
- raise errors.ParameterError(field)
+ assert False, "Declared but unhandled parameter '%s'" % field
iout.append(val)
output.append(iout)
target_node = secondary_nodes[0]
_CheckNodeOnline(self, target_node)
- # check memory requirements on the secondary node
- _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
- instance.name, bep[constants.BE_MEMORY],
- instance.hypervisor)
+ _CheckNodeNotDrained(self, target_node)
+ if instance.admin_up:
+ # check memory requirements on the secondary node
+ _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
+ instance.name, bep[constants.BE_MEMORY],
+ instance.hypervisor)
+ else:
+ self.LogInfo("Not checking memory on the secondary node as"
+ " instance will not be started")
# check bridge existance
- brlist = [nic.bridge for nic in instance.nics]
- result = self.rpc.call_bridges_exist(target_node, brlist)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("One or more target bridges %s does not"
- " exist on destination node '%s'" %
- (brlist, target_node))
+ _CheckInstanceBridgesExist(self, instance, node=target_node)
def Exec(self, feedback_fn):
"""Failover an instance.
for dev in instance.disks:
# for drbd, these are drbd over lvm
if not _CheckDiskConsistency(self, dev, target_node, False):
- if instance.status == "up" and not self.op.ignore_consistency:
+ if instance.admin_up and not self.op.ignore_consistency:
raise errors.OpExecError("Disk %s is degraded on target node,"
" aborting failover." % dev.iv_name)
instance.name, source_node)
result = self.rpc.call_instance_shutdown(source_node, instance)
- if result.failed or not result.data:
+ msg = result.fail_msg
+ if msg:
if self.op.ignore_consistency:
self.proc.LogWarning("Could not shutdown instance %s on node %s."
- " Proceeding"
- " anyway. Please make sure node %s is down",
- instance.name, source_node, source_node)
+ " Proceeding anyway. Please make sure node"
+ " %s is down. Error details: %s",
+ instance.name, source_node, source_node, msg)
else:
- raise errors.OpExecError("Could not shutdown instance %s on node %s" %
- (instance.name, source_node))
+ raise errors.OpExecError("Could not shutdown instance %s on"
+ " node %s: %s" %
+ (instance.name, source_node, msg))
feedback_fn("* deactivating the instance's disks on source node")
if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
self.cfg.Update(instance)
# Only start the instance if it's marked as up
- if instance.status == "up":
+ if instance.admin_up:
feedback_fn("* activating the instance's disks on target node")
logging.info("Starting instance %s on node %s",
instance.name, target_node)
- disks_ok, dummy = _AssembleInstanceDisks(self, instance,
+ disks_ok, _ = _AssembleInstanceDisks(self, instance,
ignore_secondaries=True)
if not disks_ok:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Can't activate the instance's disks")
feedback_fn("* starting the instance on the target node")
- result = self.rpc.call_instance_start(target_node, instance, None)
- msg = result.RemoteFailMsg()
+ result = self.rpc.call_instance_start(target_node, instance, None, None)
+ msg = result.fail_msg
if msg:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Could not start instance %s on node %s: %s" %
"""
env = _BuildInstanceHookEnvByObject(self, self.instance)
+ env["MIGRATE_LIVE"] = self.op.live
+ env["MIGRATE_CLEANUP"] = self.op.cleanup
nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
return env, nl, nl
secondary_nodes = instance.secondary_nodes
if not secondary_nodes:
- raise errors.ProgrammerError("no secondary node but using "
- "drbd8 disk template")
+ raise errors.ConfigurationError("No secondary node but using"
+ " drbd8 disk template")
i_be = self.cfg.GetClusterInfo().FillBE(instance)
instance.hypervisor)
# check bridge existance
- brlist = [nic.bridge for nic in instance.nics]
- result = self.rpc.call_bridges_exist(target_node, brlist)
- if result.failed or not result.data:
- raise errors.OpPrereqError("One or more target bridges %s does not"
- " exist on destination node '%s'" %
- (brlist, target_node))
+ _CheckInstanceBridgesExist(self, instance, node=target_node)
if not self.op.cleanup:
+ _CheckNodeNotDrained(self, target_node)
result = self.rpc.call_instance_migratable(instance.primary_node,
instance)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
- msg)
+ result.Raise("Can't migrate, please use failover", prereq=True)
self.instance = instance
self.instance.disks)
min_percent = 100
for node, nres in result.items():
- msg = nres.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Cannot resync disks on node %s: %s" %
- (node, msg))
- node_done, node_percent = nres.data[1]
+ nres.Raise("Cannot resync disks on node %s" % node)
+ node_done, node_percent = nres.payload
all_done = all_done and node_done
if node_percent is not None:
min_percent = min(min_percent, node_percent)
result = self.rpc.call_blockdev_close(node, self.instance.name,
self.instance.disks)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Cannot change disk to secondary on node %s,"
- " error %s" % (node, msg))
+ result.Raise("Cannot change disk to secondary on node %s" % node)
def _GoStandalone(self):
"""Disconnect from the network.
result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
self.instance.disks)
for node, nres in result.items():
- msg = nres.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Cannot disconnect disks node %s,"
- " error %s" % (node, msg))
+ nres.Raise("Cannot disconnect disks node %s" % node)
def _GoReconnect(self, multimaster):
"""Reconnect to the network.
self.instance.disks,
self.instance.name, multimaster)
for node, nres in result.items():
- msg = nres.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Cannot change disks config on node %s,"
- " error: %s" % (node, msg))
+ nres.Raise("Cannot change disks config on node %s" % node)
def _ExecCleanup(self):
"""Try to cleanup after a failed migration.
" a bad state)")
ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
for node, result in ins_l.items():
- result.Raise()
- if not isinstance(result.data, list):
- raise errors.OpExecError("Can't contact node '%s'" % node)
+ result.Raise("Can't contact node %s" % node)
- runningon_source = instance.name in ins_l[source_node].data
- runningon_target = instance.name in ins_l[target_node].data
+ runningon_source = instance.name in ins_l[source_node].payload
+ runningon_target = instance.name in ins_l[target_node].payload
if runningon_source and runningon_target:
raise errors.OpExecError("Instance seems to be running on two nodes,"
self.feedback_fn("* done")
+ def _RevertDiskStatus(self):
+ """Try to revert the disk status after a failed migration.
+
+ """
+ target_node = self.target_node
+ try:
+ self._EnsureSecondary(target_node)
+ self._GoStandalone()
+ self._GoReconnect(False)
+ self._WaitUntilSync()
+ except errors.OpExecError, err:
+ self.LogWarning("Migration failed and I can't reconnect the"
+ " drives: error '%s'\n"
+ "Please look and recover the instance status" %
+ str(err))
+
+ def _AbortMigration(self):
+ """Call the hypervisor code to abort a started migration.
+
+ """
+ instance = self.instance
+ target_node = self.target_node
+ migration_info = self.migration_info
+
+ abort_result = self.rpc.call_finalize_migration(target_node,
+ instance,
+ migration_info,
+ False)
+ abort_msg = abort_result.fail_msg
+ if abort_msg:
+ logging.error("Aborting migration failed on target node %s: %s" %
+ (target_node, abort_msg))
+ # Don't raise an exception here, as we stil have to try to revert the
+ # disk status, even if this step failed.
+
def _ExecMigration(self):
"""Migrate an instance.
" synchronized on target node,"
" aborting migrate." % dev.iv_name)
+ # First get the migration information from the remote node
+ result = self.rpc.call_migration_info(source_node, instance)
+ msg = result.fail_msg
+ if msg:
+ log_err = ("Failed fetching source migration information from %s: %s" %
+ (source_node, msg))
+ logging.error(log_err)
+ raise errors.OpExecError(log_err)
+
+ self.migration_info = migration_info = result.payload
+
+ # Then switch the disks to master/master mode
self._EnsureSecondary(target_node)
self._GoStandalone()
self._GoReconnect(True)
self._WaitUntilSync()
+ self.feedback_fn("* preparing %s to accept the instance" % target_node)
+ result = self.rpc.call_accept_instance(target_node,
+ instance,
+ migration_info,
+ self.nodes_ip[target_node])
+
+ msg = result.fail_msg
+ if msg:
+ logging.error("Instance pre-migration failed, trying to revert"
+ " disk status: %s", msg)
+ self._AbortMigration()
+ self._RevertDiskStatus()
+ raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
+ (instance.name, msg))
+
self.feedback_fn("* migrating instance to %s" % target_node)
time.sleep(10)
result = self.rpc.call_instance_migrate(source_node, instance,
self.nodes_ip[target_node],
self.op.live)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
logging.error("Instance migration failed, trying to revert"
" disk status: %s", msg)
- try:
- self._EnsureSecondary(target_node)
- self._GoStandalone()
- self._GoReconnect(False)
- self._WaitUntilSync()
- except errors.OpExecError, err:
- self.LogWarning("Migration failed and I can't reconnect the"
- " drives: error '%s'\n"
- "Please look and recover the instance status" %
- str(err))
-
+ self._AbortMigration()
+ self._RevertDiskStatus()
raise errors.OpExecError("Could not migrate instance %s: %s" %
(instance.name, msg))
time.sleep(10)
# distribute new instance config to the other nodes
self.cfg.Update(instance)
+ result = self.rpc.call_finalize_migration(target_node,
+ instance,
+ migration_info,
+ True)
+ msg = result.fail_msg
+ if msg:
+ logging.error("Instance migration succeeded, but finalization failed:"
+ " %s" % msg)
+ raise errors.OpExecError("Could not finalize instance migration: %s" %
+ msg)
+
self._EnsureSecondary(source_node)
self._WaitUntilSync()
self._GoStandalone()
(this will be represented as a LVM tag)
@type force_open: boolean
@param force_open: this parameter will be passes to the
- L{backend.CreateBlockDevice} function where it specifies
+ L{backend.BlockdevCreate} function where it specifies
whether we run on primary or not, and it affects both
the child assembly and the device own Open() execution
(this will be represented as a LVM tag)
@type force_open: boolean
@param force_open: this parameter will be passes to the
- L{backend.CreateBlockDevice} function where it specifies
+ L{backend.BlockdevCreate} function where it specifies
whether we run on primary or not, and it affects both
the child assembly and the device own Open() execution
lu.cfg.SetDiskID(device, node)
result = lu.rpc.call_blockdev_create(node, device, device.size,
instance.name, force_open, info)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't create block device %s on"
- " node %s for instance %s: %s" %
- (device, node, instance.name, msg))
+ result.Raise("Can't create block device %s on"
+ " node %s for instance %s" % (device, node, instance.name))
if device.physical_id is None:
- device.physical_id = result.data[1]
+ device.physical_id = result.payload
def _GenerateUniqueNames(lu, exts):
if len(secondary_nodes) != 0:
raise errors.ProgrammerError("Wrong template configuration")
- names = _GenerateUniqueNames(lu, [".disk%d" % i
+ names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
for i in range(disk_count)])
for idx, disk in enumerate(disk_info):
disk_index = idx + base_index
disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
logical_id=(vgname, names[idx]),
- iv_name="disk/%d" % disk_index)
+ iv_name="disk/%d" % disk_index,
+ mode=disk["mode"])
disks.append(disk_dev)
elif template_name == constants.DT_DRBD8:
if len(secondary_nodes) != 1:
[primary_node, remote_node] * len(disk_info), instance_name)
names = []
- for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
+ for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
for i in range(disk_count)]):
names.append(lv_prefix + "_data")
names.append(lv_prefix + "_meta")
disk["size"], names[idx*2:idx*2+2],
"disk/%d" % disk_index,
minors[idx*2], minors[idx*2+1])
+ disk_dev.mode = disk["mode"]
disks.append(disk_dev)
elif template_name == constants.DT_FILE:
if len(secondary_nodes) != 0:
iv_name="disk/%d" % disk_index,
logical_id=(file_driver,
"%s/disk%d" % (file_storage_dir,
- idx)))
+ disk_index)),
+ mode=disk["mode"])
disks.append(disk_dev)
else:
raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
- if result.failed or not result.data:
- raise errors.OpExecError("Could not connect to node '%s'" % pnode)
-
- if not result.data[0]:
- raise errors.OpExecError("Failed to create directory '%s'" %
- file_storage_dir)
+ result.Raise("Failed to create directory '%s' on"
+ " node %s: %s" % (file_storage_dir, pnode))
# Note: this needs to be kept in sync with adding of disks in
# LUSetInstanceParams
"""
logging.info("Removing block devices for instance %s", instance.name)
- result = True
+ all_result = True
for device in instance.disks:
for node, disk in device.ComputeNodeTree(instance.primary_node):
lu.cfg.SetDiskID(disk, node)
- result = lu.rpc.call_blockdev_remove(node, disk)
- if result.failed or not result.data:
- lu.proc.LogWarning("Could not remove block device %s on node %s,"
- " continuing anyway", device.iv_name, node)
- result = False
+ msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
+ if msg:
+ lu.LogWarning("Could not remove block device %s on node %s,"
+ " continuing anyway: %s", device.iv_name, node, msg)
+ all_result = False
if instance.disk_template == constants.DT_FILE:
file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
file_storage_dir)
- if result.failed or not result.data:
- logging.error("Could not remove directory '%s'", file_storage_dir)
- result = False
+ msg = result.fail_msg
+ if msg:
+ lu.LogWarning("Could not remove directory '%s' on node %s: %s",
+ file_storage_dir, instance.primary_node, msg)
+ all_result = False
- return result
+ return all_result
def _ComputeDiskSize(disk_template, disks):
hvparams)
for node in nodenames:
info = hvinfo[node]
- info.Raise()
- if not info.data or not isinstance(info.data, (tuple, list)):
- raise errors.OpPrereqError("Cannot get current information"
- " from node '%s' (%s)" % (node, info.data))
- if not info.data[0]:
- raise errors.OpPrereqError("Hypervisor parameter validation failed:"
- " %s" % info.data[1])
+ if info.offline:
+ continue
+ info.Raise("Hypervisor parameter validation failed on node %s" % node)
class LUCreateInstance(LogicalUnit):
",".join(enabled_hvs)))
# check hypervisor parameter syntax (locally)
-
- filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
+ utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
+ filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
self.op.hvparams)
hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
hv_type.CheckParameterSyntax(filled_hvp)
+ self.hv_full = filled_hvp
# fill and remember the beparams dict
- utils.CheckBEParams(self.op.beparams)
- self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
+ utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
+ self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
self.op.beparams)
#### instance parameters check
# NIC buildup
self.nics = []
- for nic in self.op.nics:
+ for idx, nic in enumerate(self.op.nics):
+ nic_mode_req = nic.get("mode", None)
+ nic_mode = nic_mode_req
+ if nic_mode is None:
+ nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
+
+ # in routed mode, for the first nic, the default ip is 'auto'
+ if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
+ default_ip_mode = constants.VALUE_AUTO
+ else:
+ default_ip_mode = constants.VALUE_NONE
+
# ip validity checks
- ip = nic.get("ip", None)
- if ip is None or ip.lower() == "none":
+ ip = nic.get("ip", default_ip_mode)
+ if ip is None or ip.lower() == constants.VALUE_NONE:
nic_ip = None
elif ip.lower() == constants.VALUE_AUTO:
nic_ip = hostname1.ip
" like a valid IP" % ip)
nic_ip = ip
+ # TODO: check the ip for uniqueness !!
+ if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
+ raise errors.OpPrereqError("Routed nic mode requires an ip address")
+
# MAC address verification
mac = nic.get("mac", constants.VALUE_AUTO)
if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
raise errors.OpPrereqError("Invalid MAC address specified: %s" %
mac)
# bridge verification
- bridge = nic.get("bridge", self.cfg.GetDefBridge())
- self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
+ bridge = nic.get("bridge", None)
+ link = nic.get("link", None)
+ if bridge and link:
+ raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
+ " at the same time")
+ elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
+ raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic")
+ elif bridge:
+ link = bridge
+
+ nicparams = {}
+ if nic_mode_req:
+ nicparams[constants.NIC_MODE] = nic_mode_req
+ if link:
+ nicparams[constants.NIC_LINK] = link
+
+ check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
+ nicparams)
+ objects.NIC.CheckParameterSyntax(check_params)
+ self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
# disk checks/pre-build
self.disks = []
"""
nics = [n.ToDict() for n in self.nics]
- ial = IAllocator(self,
+ ial = IAllocator(self.cfg, self.rpc,
mode=constants.IALLOCATOR_MODE_ALLOC,
name=self.op.instance_name,
disk_template=self.op.disk_template,
"""
env = {
- "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
- "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
- "INSTANCE_ADD_MODE": self.op.mode,
+ "ADD_MODE": self.op.mode,
}
if self.op.mode == constants.INSTANCE_IMPORT:
- env["INSTANCE_SRC_NODE"] = self.op.src_node
- env["INSTANCE_SRC_PATH"] = self.op.src_path
- env["INSTANCE_SRC_IMAGES"] = self.src_images
+ env["SRC_NODE"] = self.op.src_node
+ env["SRC_PATH"] = self.op.src_path
+ env["SRC_IMAGES"] = self.src_images
- env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
+ env.update(_BuildInstanceHookEnv(
+ name=self.op.instance_name,
primary_node=self.op.pnode,
secondary_nodes=self.secondaries,
- status=self.instance_status,
+ status=self.op.start,
os_type=self.op.os_type,
memory=self.be_full[constants.BE_MEMORY],
vcpus=self.be_full[constants.BE_VCPUS],
- nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
+ nics=_NICListToTuple(self, self.nics),
+ disk_template=self.op.disk_template,
+ disks=[(d["size"], d["mode"]) for d in self.disks],
+ bep=self.be_full,
+ hvp=self.hv_full,
+ hypervisor_name=self.op.hypervisor,
))
nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
raise errors.OpPrereqError("Cluster does not support lvm-based"
" instances")
-
if self.op.mode == constants.INSTANCE_IMPORT:
src_node = self.op.src_node
src_path = self.op.src_path
if src_node is None:
- exp_list = self.rpc.call_export_list(
- self.acquired_locks[locking.LEVEL_NODE])
+ locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
+ exp_list = self.rpc.call_export_list(locked_nodes)
found = False
for node in exp_list:
- if not exp_list[node].failed and src_path in exp_list[node].data:
+ if exp_list[node].fail_msg:
+ continue
+ if src_path in exp_list[node].payload:
found = True
self.op.src_node = src_node = node
self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
_CheckNodeOnline(self, src_node)
result = self.rpc.call_export_info(src_node, src_path)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("No export found in dir %s" % src_path)
+ result.Raise("No export or invalid export found in dir %s" % src_path)
- export_info = result.data
+ export_info = objects.SerializableConfigParser.Loads(str(result.payload))
if not export_info.has_section(constants.INISECT_EXP):
raise errors.ProgrammerError("Corrupted export config")
nic_mac_ini = 'nic%d_mac' % idx
nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
+ # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
# ip ping checks (we use the same ip that was resolved in ExpandNames)
if self.op.start and not self.op.ip_check:
raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
raise errors.OpPrereqError("IP %s of instance %s already in use" %
(self.check_ip, self.op.instance_name))
+ #### mac address generation
+ # By generating here the mac address both the allocator and the hooks get
+ # the real final mac address rather than the 'auto' or 'generate' value.
+ # There is a race condition between the generation and the instance object
+ # creation, which means that we know the mac is valid now, but we're not
+ # sure it will be when we actually add the instance. If things go bad
+ # adding the instance will abort because of a duplicate mac, and the
+ # creation job will fail.
+ for nic in self.nics:
+ if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
+ nic.mac = self.cfg.GenerateMAC()
+
#### allocator run
if self.op.iallocator is not None:
if pnode.offline:
raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
pnode.name)
+ if pnode.drained:
+ raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
+ pnode.name)
self.secondaries = []
if self.op.snode == pnode.name:
raise errors.OpPrereqError("The secondary node cannot be"
" the primary node.")
- self.secondaries.append(self.op.snode)
_CheckNodeOnline(self, self.op.snode)
+ _CheckNodeNotDrained(self, self.op.snode)
+ self.secondaries.append(self.op.snode)
nodenames = [pnode.name] + self.secondaries
self.op.hypervisor)
for node in nodenames:
info = nodeinfo[node]
- info.Raise()
- info = info.data
- if not info:
- raise errors.OpPrereqError("Cannot get current information"
- " from node '%s'" % node)
+ info.Raise("Cannot get current information from node %s" % node)
+ info = info.payload
vg_free = info.get('vg_free', None)
if not isinstance(vg_free, int):
raise errors.OpPrereqError("Can't compute free disk space on"
" node %s" % node)
- if req_size > info['vg_free']:
+ if req_size > vg_free:
raise errors.OpPrereqError("Not enough disk space on target node %s."
" %d MB available, %d MB required" %
- (node, info['vg_free'], req_size))
+ (node, vg_free, req_size))
_CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
# os verification
result = self.rpc.call_os_get(pnode.name, self.op.os_type)
- result.Raise()
- if not isinstance(result.data, objects.OS):
- raise errors.OpPrereqError("OS '%s' not in supported os list for"
- " primary node" % self.op.os_type)
-
- # bridge check on primary node
- bridges = [n.bridge for n in self.nics]
- result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("One of the target bridges '%s' does not"
- " exist on destination node '%s'" %
- (",".join(bridges), pnode.name))
+ result.Raise("OS '%s' not in supported os list for primary node %s" %
+ (self.op.os_type, pnode.name), prereq=True)
+
+ _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
# memory check on primary node
if self.op.start:
self.be_full[constants.BE_MEMORY],
self.op.hypervisor)
- if self.op.start:
- self.instance_status = 'up'
- else:
- self.instance_status = 'down'
+ self.dry_run_result = list(nodenames)
def Exec(self, feedback_fn):
"""Create and add the instance to the cluster.
instance = self.op.instance_name
pnode_name = self.pnode.name
- for nic in self.nics:
- if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
- nic.mac = self.cfg.GenerateMAC()
-
ht_kind = self.op.hypervisor
if ht_kind in constants.HTS_REQ_PORT:
network_port = self.cfg.AllocatePort()
primary_node=pnode_name,
nics=self.nics, disks=disks,
disk_template=self.op.disk_template,
- status=self.instance_status,
+ admin_up=False,
network_port=network_port,
beparams=self.op.beparams,
hvparams=self.op.hvparams,
# Declare that we don't want to remove the instance lock anymore, as we've
# added the instance to the config
del self.remove_locks[locking.LEVEL_INSTANCE]
- # Remove the temp. assignements for the instance's drbds
- self.cfg.ReleaseDRBDMinors(instance)
# Unlock all the nodes
if self.op.mode == constants.INSTANCE_IMPORT:
nodes_keep = [self.op.src_node]
if iobj.disk_template != constants.DT_DISKLESS:
if self.op.mode == constants.INSTANCE_CREATE:
feedback_fn("* running the instance OS create scripts...")
- result = self.rpc.call_instance_os_add(pnode_name, iobj)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not add os for instance %s"
- " on node %s: %s" %
- (instance, pnode_name, msg))
+ result = self.rpc.call_instance_os_add(pnode_name, iobj, False)
+ result.Raise("Could not add os for instance %s"
+ " on node %s" % (instance, pnode_name))
elif self.op.mode == constants.INSTANCE_IMPORT:
feedback_fn("* running the instance OS import scripts...")
import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
src_node, src_images,
cluster_name)
- import_result.Raise()
- for idx, result in enumerate(import_result.data):
- if not result:
- self.LogWarning("Could not import the image %s for instance"
- " %s, disk %d, on node %s" %
- (src_images[idx], instance, idx, pnode_name))
+ msg = import_result.fail_msg
+ if msg:
+ self.LogWarning("Error while importing the disk images for instance"
+ " %s on node %s: %s" % (instance, pnode_name, msg))
else:
# also checked in the prereq part
raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
% self.op.mode)
if self.op.start:
+ iobj.admin_up = True
+ self.cfg.Update(iobj)
logging.info("Starting instance %s on node %s", instance, pnode_name)
feedback_fn("* starting instance...")
- result = self.rpc.call_instance_start(pnode_name, iobj, None)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not start instance: %s" % msg)
+ result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
+ result.Raise("Could not start instance")
+
+ return list(iobj.all_nodes)
class LUConnectConsole(NoHooksLU):
node_insts = self.rpc.call_instance_list([node],
[instance.hypervisor])[node]
- node_insts.Raise()
+ node_insts.Raise("Can't get node information from %s" % node)
- if instance.name not in node_insts.data:
+ if instance.name not in node_insts.payload:
raise errors.OpExecError("Instance %s is not running." % instance.name)
logging.debug("Connecting to console of %s on %s", instance.name, node)
hyper = hypervisor.GetHypervisor(instance.hypervisor)
- console_cmd = hyper.GetShellCommandForConsole(instance)
+ cluster = self.cfg.GetClusterInfo()
+ # beparams and hvparams are passed separately, to avoid editing the
+ # instance and then saving the defaults in the instance itself.
+ hvparams = cluster.FillHV(instance)
+ beparams = cluster.FillBE(instance)
+ console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
# build ssh cmdline
return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
if not hasattr(self.op, "iallocator"):
self.op.iallocator = None
- # check for valid parameter combination
- cnt = [self.op.remote_node, self.op.iallocator].count(None)
- if self.op.mode == constants.REPLACE_DISK_CHG:
- if cnt == 2:
- raise errors.OpPrereqError("When changing the secondary either an"
- " iallocator script must be used or the"
- " new node given")
- elif cnt == 0:
- raise errors.OpPrereqError("Give either the iallocator or the new"
- " secondary, not both")
- else: # not replacing the secondary
- if cnt != 2:
- raise errors.OpPrereqError("The iallocator and new node options can"
- " be used only when changing the"
- " secondary node")
+ _DiskReplacer.CheckArguments(self.op.mode, self.op.remote_node,
+ self.op.iallocator)
def ExpandNames(self):
self._ExpandAndLockInstance()
if self.op.iallocator is not None:
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
+
elif self.op.remote_node is not None:
remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
if remote_node is None:
raise errors.OpPrereqError("Node '%s' not known" %
self.op.remote_node)
+
self.op.remote_node = remote_node
+
+ # Warning: do not remove the locking of the new secondary here
+ # unless DRBD8.AddChildren is changed to work in parallel;
+ # currently it doesn't since parallel invocations of
+ # FindUnusedMinor will conflict
self.needed_locks[locking.LEVEL_NODE] = [remote_node]
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
+
else:
self.needed_locks[locking.LEVEL_NODE] = []
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
+ self.replacer = _DiskReplacer(self, self.op.instance_name, self.op.mode,
+ self.op.iallocator, self.op.remote_node,
+ self.op.disks)
+
def DeclareLocks(self, level):
# If we're not already locking all nodes in the set we have to declare the
# instance's primary/secondary nodes.
self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
self._LockInstancesNodes()
- def _RunAllocator(self):
- """Compute a new secondary node using an IAllocator.
+ def BuildHooksEnv(self):
+ """Build hooks env.
- """
- ial = IAllocator(self,
- mode=constants.IALLOCATOR_MODE_RELOC,
- name=self.op.instance_name,
- relocate_from=[self.sec_node])
-
- ial.Run(self.op.iallocator)
-
- if not ial.success:
- raise errors.OpPrereqError("Can't compute nodes using"
- " iallocator '%s': %s" % (self.op.iallocator,
- ial.info))
- if len(ial.nodes) != ial.required_nodes:
- raise errors.OpPrereqError("iallocator '%s' returned invalid number"
- " of nodes (%s), required %s" %
- (len(ial.nodes), ial.required_nodes))
- self.op.remote_node = ial.nodes[0]
- self.LogInfo("Selected new secondary for the instance: %s",
- self.op.remote_node)
-
- def BuildHooksEnv(self):
- """Build hooks env.
-
- This runs on the master, the primary and all the secondaries.
+ This runs on the master, the primary and all the secondaries.
"""
+ instance = self.replacer.instance
env = {
"MODE": self.op.mode,
"NEW_SECONDARY": self.op.remote_node,
- "OLD_SECONDARY": self.instance.secondary_nodes[0],
+ "OLD_SECONDARY": instance.secondary_nodes[0],
}
- env.update(_BuildInstanceHookEnvByObject(self, self.instance))
+ env.update(_BuildInstanceHookEnvByObject(self, instance))
nl = [
self.cfg.GetMasterNode(),
- self.instance.primary_node,
+ instance.primary_node,
]
if self.op.remote_node is not None:
nl.append(self.op.remote_node)
This checks that the instance is in the cluster.
"""
- instance = self.cfg.GetInstanceInfo(self.op.instance_name)
- assert instance is not None, \
- "Cannot retrieve locked instance %s" % self.op.instance_name
- self.instance = instance
+ self.replacer.CheckPrereq()
- if instance.disk_template != constants.DT_DRBD8:
+ def Exec(self, feedback_fn):
+ """Execute disk replacement.
+
+ This dispatches the disk replacement to the appropriate handler.
+
+ """
+ self.replacer.Exec()
+
+
+class _DiskReplacer:
+ """Replaces disks for an instance.
+
+ Note: Locking is not within the scope of this class.
+
+ """
+ def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
+ disks):
+ """Initializes this class.
+
+ """
+ # Parameters
+ self.lu = lu
+ self.instance_name = instance_name
+ self.mode = mode
+ self.iallocator_name = iallocator_name
+ self.remote_node = remote_node
+ self.disks = disks
+
+ # Shortcuts
+ self.cfg = lu.cfg
+ self.rpc = lu.rpc
+
+ # Runtime data
+ self.instance = None
+ self.new_node = None
+ self.target_node = None
+ self.other_node = None
+ self.remote_node_info = None
+ self.node_secondary_ip = None
+
+ @staticmethod
+ def CheckArguments(mode, remote_node, iallocator):
+ # check for valid parameter combination
+ cnt = [remote_node, iallocator].count(None)
+ if mode == constants.REPLACE_DISK_CHG:
+ if cnt == 2:
+ raise errors.OpPrereqError("When changing the secondary either an"
+ " iallocator script must be used or the"
+ " new node given")
+ elif cnt == 0:
+ raise errors.OpPrereqError("Give either the iallocator or the new"
+ " secondary, not both")
+ else: # not replacing the secondary
+ if cnt != 2:
+ raise errors.OpPrereqError("The iallocator and new node options can"
+ " be used only when changing the"
+ " secondary node")
+
+ @staticmethod
+ def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
+ """Compute a new secondary node using an IAllocator.
+
+ """
+ ial = IAllocator(lu.cfg, lu.rpc,
+ mode=constants.IALLOCATOR_MODE_RELOC,
+ name=instance_name,
+ relocate_from=relocate_from)
+
+ ial.Run(iallocator_name)
+
+ if not ial.success:
+ raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
+ " %s" % (iallocator_name, ial.info))
+
+ if len(ial.nodes) != ial.required_nodes:
+ raise errors.OpPrereqError("iallocator '%s' returned invalid number"
+ " of nodes (%s), required %s" %
+ (len(ial.nodes), ial.required_nodes))
+
+ remote_node_name = ial.nodes[0]
+
+ lu.LogInfo("Selected new secondary for instance '%s': %s",
+ instance_name, remote_node_name)
+
+ return remote_node_name
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This checks that the instance is in the cluster.
+
+ """
+ self.instance = self.cfg.GetInstanceInfo(self.instance_name)
+ assert self.instance is not None, \
+ "Cannot retrieve locked instance %s" % self.instance_name
+
+ if self.instance.disk_template != constants.DT_DRBD8:
raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
" instances")
- if len(instance.secondary_nodes) != 1:
+ if len(self.instance.secondary_nodes) != 1:
raise errors.OpPrereqError("The instance has a strange layout,"
" expected one secondary but found %d" %
- len(instance.secondary_nodes))
+ len(self.instance.secondary_nodes))
- self.sec_node = instance.secondary_nodes[0]
+ secondary_node = self.instance.secondary_nodes[0]
- if self.op.iallocator is not None:
- self._RunAllocator()
+ if self.iallocator_name is None:
+ remote_node = self.remote_node
+ else:
+ remote_node = self._RunAllocator(self.lu, self.iallocator_name,
+ self.instance.name, secondary_node)
- remote_node = self.op.remote_node
if remote_node is not None:
self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
assert self.remote_node_info is not None, \
"Cannot retrieve locked node %s" % remote_node
else:
self.remote_node_info = None
- if remote_node == instance.primary_node:
+
+ if remote_node == self.instance.primary_node:
raise errors.OpPrereqError("The specified node is the primary node of"
" the instance.")
- elif remote_node == self.sec_node:
+
+ if remote_node == secondary_node:
raise errors.OpPrereqError("The specified node is already the"
" secondary node of the instance.")
- if self.op.mode == constants.REPLACE_DISK_PRI:
- n1 = self.tgt_node = instance.primary_node
- n2 = self.oth_node = self.sec_node
- elif self.op.mode == constants.REPLACE_DISK_SEC:
- n1 = self.tgt_node = self.sec_node
- n2 = self.oth_node = instance.primary_node
- elif self.op.mode == constants.REPLACE_DISK_CHG:
- n1 = self.new_node = remote_node
- n2 = self.oth_node = instance.primary_node
- self.tgt_node = self.sec_node
+ if self.mode == constants.REPLACE_DISK_PRI:
+ self.target_node = self.instance.primary_node
+ self.other_node = secondary_node
+ check_nodes = [self.target_node, self.other_node]
+
+ elif self.mode == constants.REPLACE_DISK_SEC:
+ self.target_node = secondary_node
+ self.other_node = self.instance.primary_node
+ check_nodes = [self.target_node, self.other_node]
+
+ elif self.mode == constants.REPLACE_DISK_CHG:
+ self.new_node = remote_node
+ self.other_node = self.instance.primary_node
+ self.target_node = secondary_node
+ check_nodes = [self.new_node, self.other_node]
+
+ _CheckNodeNotDrained(self.lu, remote_node)
+
else:
- raise errors.ProgrammerError("Unhandled disk replace mode")
+ raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
+ self.mode)
+
+ for node in check_nodes:
+ _CheckNodeOnline(self.lu, node)
+
+ # If not specified all disks should be replaced
+ if not self.disks:
+ self.disks = range(len(self.instance.disks))
+
+ # Check whether disks are valid
+ for disk_idx in self.disks:
+ self.instance.FindDisk(disk_idx)
+
+ # Get secondary node IP addresses
+ node_2nd_ip = {}
+
+ for node_name in [self.target_node, self.other_node, self.new_node]:
+ if node_name is not None:
+ node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
+
+ self.node_secondary_ip = node_2nd_ip
+
+ def Exec(self):
+ """Execute disk replacement.
+
+ This dispatches the disk replacement to the appropriate handler.
- _CheckNodeOnline(self, n1)
- _CheckNodeOnline(self, n2)
+ """
+ activate_disks = (not self.instance.admin_up)
+
+ # Activate the instance disks if we're replacing them on a down instance
+ if activate_disks:
+ _StartInstanceDisks(self.lu, self.instance, True)
+
+ try:
+ if self.mode == constants.REPLACE_DISK_CHG:
+ return self._ExecDrbd8Secondary()
+ else:
+ return self._ExecDrbd8DiskOnly()
+
+ finally:
+ # Deactivate the instance disks if we're replacing them on a down instance
+ if activate_disks:
+ _SafeShutdownInstanceDisks(self.lu, self.instance)
+
+ def _CheckVolumeGroup(self, nodes):
+ self.lu.LogInfo("Checking volume groups")
- if not self.op.disks:
- self.op.disks = range(len(instance.disks))
+ vgname = self.cfg.GetVGName()
+
+ # Make sure volume group exists on all involved nodes
+ results = self.rpc.call_vg_list(nodes)
+ if not results:
+ raise errors.OpExecError("Can't list volume groups on the nodes")
+
+ for node in nodes:
+ res = results[node]
+ res.Raise("Error checking node %s" % node)
+ if vgname not in res.payload:
+ raise errors.OpExecError("Volume group '%s' not found on node %s" %
+ (vgname, node))
+
+ def _CheckDisksExistence(self, nodes):
+ # Check disk existence
+ for idx, dev in enumerate(self.instance.disks):
+ if idx not in self.disks:
+ continue
- for disk_idx in self.op.disks:
- instance.FindDisk(disk_idx)
+ for node in nodes:
+ self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
+ self.cfg.SetDiskID(dev, node)
- def _ExecD8DiskOnly(self, feedback_fn):
- """Replace a disk on the primary or secondary for dbrd8.
+ result = self.rpc.call_blockdev_find(node, dev)
+
+ msg = result.fail_msg
+ if msg or not result.payload:
+ if not msg:
+ msg = "disk not found"
+ raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
+ (idx, node, msg))
+
+ def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
+ for idx, dev in enumerate(self.instance.disks):
+ if idx not in self.disks:
+ continue
+
+ self.lu.LogInfo("Checking disk/%d consistency on node %s" %
+ (idx, node_name))
+
+ if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
+ ldisk=ldisk):
+ raise errors.OpExecError("Node %s has degraded storage, unsafe to"
+ " replace disks for instance %s" %
+ (node_name, self.instance.name))
+
+ def _CreateNewStorage(self, node_name):
+ vgname = self.cfg.GetVGName()
+ iv_names = {}
+
+ for idx, dev in enumerate(self.instance.disks):
+ if idx not in self.disks:
+ continue
+
+ self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
+
+ self.cfg.SetDiskID(dev, node_name)
+
+ lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
+ names = _GenerateUniqueNames(self.lu, lv_names)
+
+ lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
+ logical_id=(vgname, names[0]))
+ lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
+ logical_id=(vgname, names[1]))
+
+ new_lvs = [lv_data, lv_meta]
+ old_lvs = dev.children
+ iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
+
+ # we pass force_create=True to force the LVM creation
+ for new_lv in new_lvs:
+ _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
+ _GetInstanceInfoText(self.instance), False)
+
+ return iv_names
+
+ def _CheckDevices(self, node_name, iv_names):
+ for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
+ self.cfg.SetDiskID(dev, node_name)
+
+ result = self.rpc.call_blockdev_find(node_name, dev)
+
+ msg = result.fail_msg
+ if msg or not result.payload:
+ if not msg:
+ msg = "disk not found"
+ raise errors.OpExecError("Can't find DRBD device %s: %s" %
+ (name, msg))
+
+ if result.payload[5]:
+ raise errors.OpExecError("DRBD device %s is degraded!" % name)
+
+ def _RemoveOldStorage(self, node_name, iv_names):
+ for name, (dev, old_lvs, _) in iv_names.iteritems():
+ self.lu.LogInfo("Remove logical volumes for %s" % name)
+
+ for lv in old_lvs:
+ self.cfg.SetDiskID(lv, node_name)
+
+ msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
+ if msg:
+ self.lu.LogWarning("Can't remove old LV: %s" % msg,
+ hint="remove unused LVs manually")
+
+ def _ExecDrbd8DiskOnly(self):
+ """Replace a disk on the primary or secondary for DRBD 8.
The algorithm for replace is quite complicated:
"""
steps_total = 6
- warning, info = (self.proc.LogWarning, self.proc.LogInfo)
- instance = self.instance
- iv_names = {}
- vgname = self.cfg.GetVGName()
- # start of work
- cfg = self.cfg
- tgt_node = self.tgt_node
- oth_node = self.oth_node
# Step: check device activation
- self.proc.LogStep(1, steps_total, "check device existence")
- info("checking volume groups")
- my_vg = cfg.GetVGName()
- results = self.rpc.call_vg_list([oth_node, tgt_node])
- if not results:
- raise errors.OpExecError("Can't list volume groups on the nodes")
- for node in oth_node, tgt_node:
- res = results[node]
- if res.failed or not res.data or my_vg not in res.data:
- raise errors.OpExecError("Volume group '%s' not found on %s" %
- (my_vg, node))
- for idx, dev in enumerate(instance.disks):
- if idx not in self.op.disks:
- continue
- for node in tgt_node, oth_node:
- info("checking disk/%d on %s" % (idx, node))
- cfg.SetDiskID(dev, node)
- if not self.rpc.call_blockdev_find(node, dev):
- raise errors.OpExecError("Can't find disk/%d on node %s" %
- (idx, node))
+ self.lu.LogStep(1, steps_total, "Check device existence")
+ self._CheckDisksExistence([self.other_node, self.target_node])
+ self._CheckVolumeGroup([self.target_node, self.other_node])
# Step: check other node consistency
- self.proc.LogStep(2, steps_total, "check peer consistency")
- for idx, dev in enumerate(instance.disks):
- if idx not in self.op.disks:
- continue
- info("checking disk/%d consistency on %s" % (idx, oth_node))
- if not _CheckDiskConsistency(self, dev, oth_node,
- oth_node==instance.primary_node):
- raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
- " to replace disks on this node (%s)" %
- (oth_node, tgt_node))
+ self.lu.LogStep(2, steps_total, "Check peer consistency")
+ self._CheckDisksConsistency(self.other_node,
+ self.other_node == self.instance.primary_node,
+ False)
# Step: create new storage
- self.proc.LogStep(3, steps_total, "allocate new storage")
- for idx, dev in enumerate(instance.disks):
- if idx not in self.op.disks:
- continue
- size = dev.size
- cfg.SetDiskID(dev, tgt_node)
- lv_names = [".disk%d_%s" % (idx, suf)
- for suf in ["data", "meta"]]
- names = _GenerateUniqueNames(self, lv_names)
- lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
- logical_id=(vgname, names[0]))
- lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
- logical_id=(vgname, names[1]))
- new_lvs = [lv_data, lv_meta]
- old_lvs = dev.children
- iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
- info("creating new local storage on %s for %s" %
- (tgt_node, dev.iv_name))
- # we pass force_create=True to force the LVM creation
- for new_lv in new_lvs:
- _CreateBlockDev(self, tgt_node, instance, new_lv, True,
- _GetInstanceInfoText(instance), False)
+ self.lu.LogStep(3, steps_total, "Allocate new storage")
+ iv_names = self._CreateNewStorage(self.target_node)
# Step: for each lv, detach+rename*2+attach
- self.proc.LogStep(4, steps_total, "change drbd configuration")
+ self.lu.LogStep(4, steps_total, "Changing drbd configuration")
for dev, old_lvs, new_lvs in iv_names.itervalues():
- info("detaching %s drbd from local storage" % dev.iv_name)
- result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Can't detach drbd from local storage on node"
- " %s for device %s" % (tgt_node, dev.iv_name))
+ self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
+
+ result = self.rpc.call_blockdev_removechildren(self.target_node, dev, old_lvs)
+ result.Raise("Can't detach drbd from local storage on node"
+ " %s for device %s" % (self.target_node, dev.iv_name))
#dev.children = []
#cfg.Update(instance)
temp_suffix = int(time.time())
ren_fn = lambda d, suff: (d.physical_id[0],
d.physical_id[1] + "_replaced-%s" % suff)
- # build the rename list based on what LVs exist on the node
- rlist = []
+
+ # Build the rename list based on what LVs exist on the node
+ rename_old_to_new = []
for to_ren in old_lvs:
- find_res = self.rpc.call_blockdev_find(tgt_node, to_ren)
- if not find_res.failed and find_res.data is not None: # device exists
- rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
-
- info("renaming the old LVs on the target node")
- result = self.rpc.call_blockdev_rename(tgt_node, rlist)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
- # now we rename the new LVs to the old LVs
- info("renaming the new LVs on the target node")
- rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
- result = self.rpc.call_blockdev_rename(tgt_node, rlist)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
+ result = self.rpc.call_blockdev_find(self.target_node, to_ren)
+ if not result.fail_msg and result.payload:
+ # device exists
+ rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
+
+ self.lu.LogInfo("Renaming the old LVs on the target node")
+ result = self.rpc.call_blockdev_rename(self.target_node, rename_old_to_new)
+ result.Raise("Can't rename old LVs on node %s" % self.target_node)
+
+ # Now we rename the new LVs to the old LVs
+ self.lu.LogInfo("Renaming the new LVs on the target node")
+ rename_new_to_old = [(new, old.physical_id)
+ for old, new in zip(old_lvs, new_lvs)]
+ result = self.rpc.call_blockdev_rename(self.target_node, rename_new_to_old)
+ result.Raise("Can't rename new LVs on node %s" % self.target_node)
for old, new in zip(old_lvs, new_lvs):
new.logical_id = old.logical_id
- cfg.SetDiskID(new, tgt_node)
+ self.cfg.SetDiskID(new, self.target_node)
for disk in old_lvs:
disk.logical_id = ren_fn(disk, temp_suffix)
- cfg.SetDiskID(disk, tgt_node)
+ self.cfg.SetDiskID(disk, self.target_node)
- # now that the new lvs have the old name, we can add them to the device
- info("adding new mirror component on %s" % tgt_node)
- result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
- if result.failed or not result.data:
+ # Now that the new lvs have the old name, we can add them to the device
+ self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
+ result = self.rpc.call_blockdev_addchildren(self.target_node, dev, new_lvs)
+ msg = result.fail_msg
+ if msg:
for new_lv in new_lvs:
- result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
- if result.failed or not result.data:
- warning("Can't rollback device %s", hint="manually cleanup unused"
- " logical volumes")
- raise errors.OpExecError("Can't add local storage to drbd")
+ msg2 = self.rpc.call_blockdev_remove(self.target_node, new_lv).fail_msg
+ if msg2:
+ self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
+ hint=("cleanup manually the unused logical"
+ "volumes"))
+ raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
dev.children = new_lvs
- cfg.Update(instance)
- # Step: wait for sync
+ self.cfg.Update(self.instance)
- # this can fail as the old devices are degraded and _WaitForSync
- # does a combined result over all disks, so we don't check its
- # return value
- self.proc.LogStep(5, steps_total, "sync devices")
- _WaitForSync(self, instance, unlock=True)
+ # Wait for sync
+ # This can fail as the old devices are degraded and _WaitForSync
+ # does a combined result over all disks, so we don't check its return value
+ self.lu.LogStep(5, steps_total, "Sync devices")
+ _WaitForSync(self.lu, self.instance, unlock=True)
- # so check manually all the devices
- for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
- cfg.SetDiskID(dev, instance.primary_node)
- result = self.rpc.call_blockdev_find(instance.primary_node, dev)
- if result.failed or result.data[5]:
- raise errors.OpExecError("DRBD device %s is degraded!" % name)
+ # Check all devices manually
+ self._CheckDevices(self.instance.primary_node, iv_names)
# Step: remove old storage
- self.proc.LogStep(6, steps_total, "removing old storage")
- for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
- info("remove logical volumes for %s" % name)
- for lv in old_lvs:
- cfg.SetDiskID(lv, tgt_node)
- result = self.rpc.call_blockdev_remove(tgt_node, lv)
- if result.failed or not result.data:
- warning("Can't remove old LV", hint="manually remove unused LVs")
- continue
+ self.lu.LogStep(6, steps_total, "Removing old storage")
+ self._RemoveOldStorage(self.target_node, iv_names)
- def _ExecD8Secondary(self, feedback_fn):
- """Replace the secondary node for drbd8.
+ def _ExecDrbd8Secondary(self):
+ """Replace the secondary node for DRBD 8.
The algorithm for replace is quite complicated:
- for all disks of the instance:
"""
steps_total = 6
- warning, info = (self.proc.LogWarning, self.proc.LogInfo)
- instance = self.instance
- iv_names = {}
- # start of work
- cfg = self.cfg
- old_node = self.tgt_node
- new_node = self.new_node
- pri_node = instance.primary_node
- nodes_ip = {
- old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
- new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
- pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
- }
# Step: check device activation
- self.proc.LogStep(1, steps_total, "check device existence")
- info("checking volume groups")
- my_vg = cfg.GetVGName()
- results = self.rpc.call_vg_list([pri_node, new_node])
- for node in pri_node, new_node:
- res = results[node]
- if res.failed or not res.data or my_vg not in res.data:
- raise errors.OpExecError("Volume group '%s' not found on %s" %
- (my_vg, node))
- for idx, dev in enumerate(instance.disks):
- if idx not in self.op.disks:
- continue
- info("checking disk/%d on %s" % (idx, pri_node))
- cfg.SetDiskID(dev, pri_node)
- result = self.rpc.call_blockdev_find(pri_node, dev)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Can't find disk/%d on node %s" %
- (idx, pri_node))
+ self.lu.LogStep(1, steps_total, "Check device existence")
+ self._CheckDisksExistence([self.instance.primary_node])
+ self._CheckVolumeGroup([self.instance.primary_node])
# Step: check other node consistency
- self.proc.LogStep(2, steps_total, "check peer consistency")
- for idx, dev in enumerate(instance.disks):
- if idx not in self.op.disks:
- continue
- info("checking disk/%d consistency on %s" % (idx, pri_node))
- if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
- raise errors.OpExecError("Primary node (%s) has degraded storage,"
- " unsafe to replace the secondary" %
- pri_node)
+ self.lu.LogStep(2, steps_total, "Check peer consistency")
+ self._CheckDisksConsistency(self.instance.primary_node, True, True)
# Step: create new storage
- self.proc.LogStep(3, steps_total, "allocate new storage")
- for idx, dev in enumerate(instance.disks):
- info("adding new local storage on %s for disk/%d" %
- (new_node, idx))
+ self.lu.LogStep(3, steps_total, "Allocate new storage")
+ for idx, dev in enumerate(self.instance.disks):
+ self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
+ (self.new_node, idx))
# we pass force_create=True to force LVM creation
for new_lv in dev.children:
- _CreateBlockDev(self, new_node, instance, new_lv, True,
- _GetInstanceInfoText(instance), False)
+ _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
+ _GetInstanceInfoText(self.instance), False)
# Step 4: dbrd minors and drbd setups changes
# after this, we must manually remove the drbd minors on both the
# error and the success paths
- minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
- instance.name)
- logging.debug("Allocated minors %s" % (minors,))
- self.proc.LogStep(4, steps_total, "changing drbd configuration")
- for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
- size = dev.size
- info("activating a new drbd on %s for disk/%d" % (new_node, idx))
+ self.lu.LogStep(4, steps_total, "Changing drbd configuration")
+ minors = self.cfg.AllocateDRBDMinor([self.new_node for dev in self.instance.disks],
+ self.instance.name)
+ logging.debug("Allocated minors %r" % (minors,))
+
+ iv_names = {}
+ for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
+ self.lu.LogInfo("activating a new drbd on %s for disk/%d" % (self.new_node, idx))
# create new devices on new_node; note that we create two IDs:
# one without port, so the drbd will be activated without
# networking information on the new node at this stage, and one
# with network, for the latter activation in step 4
(o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
- if pri_node == o_node1:
+ if self.instance.primary_node == o_node1:
p_minor = o_minor1
else:
p_minor = o_minor2
- new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
- new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
+ new_alone_id = (self.instance.primary_node, self.new_node, None, p_minor, new_minor, o_secret)
+ new_net_id = (self.instance.primary_node, self.new_node, o_port, p_minor, new_minor, o_secret)
iv_names[idx] = (dev, dev.children, new_net_id)
logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
new_net_id)
new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
logical_id=new_alone_id,
- children=dev.children)
+ children=dev.children,
+ size=dev.size)
try:
- _CreateSingleBlockDev(self, new_node, instance, new_drbd,
- _GetInstanceInfoText(instance), False)
- except errors.BlockDeviceError:
- self.cfg.ReleaseDRBDMinors(instance.name)
+ _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
+ _GetInstanceInfoText(self.instance), False)
+ except errors.GenericError:
+ self.cfg.ReleaseDRBDMinors(self.instance.name)
raise
- for idx, dev in enumerate(instance.disks):
- # we have new devices, shutdown the drbd on the old secondary
- info("shutting down drbd for disk/%d on old node" % idx)
- cfg.SetDiskID(dev, old_node)
- result = self.rpc.call_blockdev_shutdown(old_node, dev)
- if result.failed or not result.data:
- warning("Failed to shutdown drbd for disk/%d on old node" % idx,
- hint="Please cleanup this device manually as soon as possible")
+ # We have new devices, shutdown the drbd on the old secondary
+ for idx, dev in enumerate(self.instance.disks):
+ self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
+ self.cfg.SetDiskID(dev, self.target_node)
+ msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
+ if msg:
+ self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
+ "node: %s" % (idx, msg),
+ hint=("Please cleanup this device manually as"
+ " soon as possible"))
- info("detaching primary drbds from the network (=> standalone)")
- result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
- instance.disks)[pri_node]
+ self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
+ result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node], self.node_secondary_ip,
+ self.instance.disks)[self.instance.primary_node]
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
# detaches didn't succeed (unlikely)
- self.cfg.ReleaseDRBDMinors(instance.name)
+ self.cfg.ReleaseDRBDMinors(self.instance.name)
raise errors.OpExecError("Can't detach the disks from the network on"
" old node: %s" % (msg,))
# if we managed to detach at least one, we update all the disks of
# the instance to point to the new secondary
- info("updating instance configuration")
+ self.lu.LogInfo("Updating instance configuration")
for dev, _, new_logical_id in iv_names.itervalues():
dev.logical_id = new_logical_id
- cfg.SetDiskID(dev, pri_node)
- cfg.Update(instance)
- # we can remove now the temp minors as now the new values are
- # written to the config file (and therefore stable)
- self.cfg.ReleaseDRBDMinors(instance.name)
+ self.cfg.SetDiskID(dev, self.instance.primary_node)
+
+ self.cfg.Update(self.instance)
# and now perform the drbd attach
- info("attaching primary drbds to new secondary (standalone => connected)")
- result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
- instance.disks, instance.name,
+ self.lu.LogInfo("Attaching primary drbds to new secondary"
+ " (standalone => connected)")
+ result = self.rpc.call_drbd_attach_net([self.instance.primary_node, self.new_node], self.node_secondary_ip,
+ self.instance.disks, self.instance.name,
False)
for to_node, to_result in result.items():
- msg = to_result.RemoteFailMsg()
+ msg = to_result.fail_msg
if msg:
- warning("can't attach drbd disks on node %s: %s", to_node, msg,
- hint="please do a gnt-instance info to see the"
- " status of disks")
-
- # this can fail as the old devices are degraded and _WaitForSync
- # does a combined result over all disks, so we don't check its
- # return value
- self.proc.LogStep(5, steps_total, "sync devices")
- _WaitForSync(self, instance, unlock=True)
-
- # so check manually all the devices
- for idx, (dev, old_lvs, _) in iv_names.iteritems():
- cfg.SetDiskID(dev, pri_node)
- result = self.rpc.call_blockdev_find(pri_node, dev)
- result.Raise()
- if result.data[5]:
- raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
-
- self.proc.LogStep(6, steps_total, "removing old storage")
- for idx, (dev, old_lvs, _) in iv_names.iteritems():
- info("remove logical volumes for disk/%d" % idx)
- for lv in old_lvs:
- cfg.SetDiskID(lv, old_node)
- result = self.rpc.call_blockdev_remove(old_node, lv)
- if result.failed or not result.data:
- warning("Can't remove LV on old secondary",
- hint="Cleanup stale volumes by hand")
-
- def Exec(self, feedback_fn):
- """Execute disk replacement.
+ self.lu.LogWarning("Can't attach drbd disks on node %s: %s", to_node, msg,
+ hint=("please do a gnt-instance info to see the"
+ " status of disks"))
- This dispatches the disk replacement to the appropriate handler.
+ # Wait for sync
+ # This can fail as the old devices are degraded and _WaitForSync
+ # does a combined result over all disks, so we don't check its return value
+ self.lu.LogStep(5, steps_total, "Sync devices")
+ _WaitForSync(self.lu, self.instance, unlock=True)
- """
- instance = self.instance
+ # Check all devices manually
+ self._CheckDevices(self.instance.primary_node, iv_names)
- # Activate the instance disks if we're replacing them on a down instance
- if instance.status == "down":
- _StartInstanceDisks(self, instance, True)
-
- if self.op.mode == constants.REPLACE_DISK_CHG:
- fn = self._ExecD8Secondary
- else:
- fn = self._ExecD8DiskOnly
-
- ret = fn(feedback_fn)
-
- # Deactivate the instance disks if we're replacing them on a down instance
- if instance.status == "down":
- _SafeShutdownInstanceDisks(self, instance)
-
- return ret
+ # Step: remove old storage
+ self.lu.LogStep(6, steps_total, "Removing old storage")
+ self._RemoveOldStorage(self.target_node, iv_names)
class LUGrowDisk(LogicalUnit):
instance.hypervisor)
for node in nodenames:
info = nodeinfo[node]
- if info.failed or not info.data:
- raise errors.OpPrereqError("Cannot get current information"
- " from node '%s'" % node)
- vg_free = info.data.get('vg_free', None)
+ info.Raise("Cannot get current information from node %s" % node)
+ vg_free = info.payload.get('vg_free', None)
if not isinstance(vg_free, int):
raise errors.OpPrereqError("Can't compute free disk space on"
" node %s" % node)
for node in instance.all_nodes:
self.cfg.SetDiskID(disk, node)
result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
- result.Raise()
- if (not result.data or not isinstance(result.data, (list, tuple)) or
- len(result.data) != 2):
- raise errors.OpExecError("Grow request failed to node %s" % node)
- elif not result.data[0]:
- raise errors.OpExecError("Grow request failed to node %s: %s" %
- (node, result.data[1]))
+ result.Raise("Grow request failed to node %s" % node)
disk.RecordGrow(self.op.amount)
self.cfg.Update(instance)
if self.op.wait_for_sync:
def ExpandNames(self):
self.needed_locks = {}
- self.share_locks = dict(((i, 1) for i in locking.LEVELS))
+ self.share_locks = dict.fromkeys(locking.LEVELS, 1)
if not isinstance(self.op.instances, list):
raise errors.OpPrereqError("Invalid argument type 'instances'")
if not static:
self.cfg.SetDiskID(dev, instance.primary_node)
dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
- dev_pstatus.Raise()
- dev_pstatus = dev_pstatus.data
+ if dev_pstatus.offline:
+ dev_pstatus = None
+ else:
+ dev_pstatus.Raise("Can't compute disk status for %s" % instance.name)
+ dev_pstatus = dev_pstatus.payload
else:
dev_pstatus = None
if snode and not static:
self.cfg.SetDiskID(dev, snode)
dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
- dev_sstatus.Raise()
- dev_sstatus = dev_sstatus.data
+ if dev_sstatus.offline:
+ dev_sstatus = None
+ else:
+ dev_sstatus.Raise("Can't compute disk status for %s" % instance.name)
+ dev_sstatus = dev_sstatus.payload
else:
dev_sstatus = None
"sstatus": dev_sstatus,
"children": dev_children,
"mode": dev.mode,
+ "size": dev.size,
}
return data
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- remote_info.Raise()
- remote_info = remote_info.data
+ remote_info.Raise("Error checking node %s" % instance.primary_node)
+ remote_info = remote_info.payload
if remote_info and "state" in remote_info:
remote_state = "up"
else:
remote_state = "down"
else:
remote_state = None
- if instance.status == "down":
- config_state = "down"
- else:
+ if instance.admin_up:
config_state = "up"
+ else:
+ config_state = "down"
disks = [self._ComputeDiskStatus(instance, None, device)
for device in instance.disks]
"pnode": instance.primary_node,
"snodes": instance.secondary_nodes,
"os": instance.os,
- "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
+ # this happens to be the same format used for hooks
+ "nics": _NICListToTuple(self, instance.nics),
"disks": disks,
"hypervisor": instance.hypervisor,
"network_port": instance.network_port,
self.op.hvparams or self.op.beparams):
raise errors.OpPrereqError("No changes submitted")
- utils.CheckBEParams(self.op.beparams)
-
# Disk validation
disk_addremove = 0
for disk_op, disk_dict in self.op.disks:
else:
if not isinstance(disk_op, int):
raise errors.OpPrereqError("Invalid disk index")
+ if not isinstance(disk_dict, dict):
+ msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
+ raise errors.OpPrereqError(msg)
+
if disk_op == constants.DDM_ADD:
mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
- if mode not in (constants.DISK_RDONLY, constants.DISK_RDWR):
+ if mode not in constants.DISK_ACCESS_SET:
raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
size = disk_dict.get('size', None)
if size is None:
else:
if not isinstance(nic_op, int):
raise errors.OpPrereqError("Invalid nic index")
+ if not isinstance(nic_dict, dict):
+ msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
+ raise errors.OpPrereqError(msg)
# nic_dict should be a dict
nic_ip = nic_dict.get('ip', None)
if nic_ip is not None:
- if nic_ip.lower() == "none":
+ if nic_ip.lower() == constants.VALUE_NONE:
nic_dict['ip'] = None
else:
if not utils.IsValidIP(nic_ip):
raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
- # we can only check None bridges and assign the default one
+
nic_bridge = nic_dict.get('bridge', None)
- if nic_bridge is None:
- nic_dict['bridge'] = self.cfg.GetDefBridge()
- # but we can validate MACs
- nic_mac = nic_dict.get('mac', None)
- if nic_mac is not None:
- if self.cfg.IsMacInUse(nic_mac):
- raise errors.OpPrereqError("MAC address %s already in use"
- " in cluster" % nic_mac)
+ nic_link = nic_dict.get('link', None)
+ if nic_bridge and nic_link:
+ raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
+ " at the same time")
+ elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
+ nic_dict['bridge'] = None
+ elif nic_link and nic_link.lower() == constants.VALUE_NONE:
+ nic_dict['link'] = None
+
+ if nic_op == constants.DDM_ADD:
+ nic_mac = nic_dict.get('mac', None)
+ if nic_mac is None:
+ nic_dict['mac'] = constants.VALUE_AUTO
+
+ if 'mac' in nic_dict:
+ nic_mac = nic_dict['mac']
if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
if not utils.IsValidMac(nic_mac):
raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
+ if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
+ raise errors.OpPrereqError("'auto' is not a valid MAC address when"
+ " modifying an existing nic")
+
if nic_addremove > 1:
raise errors.OpPrereqError("Only one NIC add or remove operation"
" supported at a time")
args['memory'] = self.be_new[constants.BE_MEMORY]
if constants.BE_VCPUS in self.be_new:
args['vcpus'] = self.be_new[constants.BE_VCPUS]
- # FIXME: readd disk/nic changes
+ # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
+ # information at all.
+ if self.op.nics:
+ args['nics'] = []
+ nic_override = dict(self.op.nics)
+ c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
+ for idx, nic in enumerate(self.instance.nics):
+ if idx in nic_override:
+ this_nic_override = nic_override[idx]
+ else:
+ this_nic_override = {}
+ if 'ip' in this_nic_override:
+ ip = this_nic_override['ip']
+ else:
+ ip = nic.ip
+ if 'mac' in this_nic_override:
+ mac = this_nic_override['mac']
+ else:
+ mac = nic.mac
+ if idx in self.nic_pnew:
+ nicparams = self.nic_pnew[idx]
+ else:
+ nicparams = objects.FillDict(c_nicparams, nic.nicparams)
+ mode = nicparams[constants.NIC_MODE]
+ link = nicparams[constants.NIC_LINK]
+ args['nics'].append((ip, mac, mode, link))
+ if constants.DDM_ADD in nic_override:
+ ip = nic_override[constants.DDM_ADD].get('ip', None)
+ mac = nic_override[constants.DDM_ADD]['mac']
+ nicparams = self.nic_pnew[constants.DDM_ADD]
+ mode = nicparams[constants.NIC_MODE]
+ link = nicparams[constants.NIC_LINK]
+ args['nics'].append((ip, mac, mode, link))
+ elif constants.DDM_REMOVE in nic_override:
+ del args['nics'][-1]
+
env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
return env, nl, nl
+ def _GetUpdatedParams(self, old_params, update_dict,
+ default_values, parameter_types):
+ """Return the new params dict for the given params.
+
+ @type old_params: dict
+ @param old_params: old parameters
+ @type update_dict: dict
+ @param update_dict: dict containing new parameter values,
+ or constants.VALUE_DEFAULT to reset the
+ parameter to its default value
+ @type default_values: dict
+ @param default_values: default values for the filled parameters
+ @type parameter_types: dict
+ @param parameter_types: dict mapping target dict keys to types
+ in constants.ENFORCEABLE_TYPES
+ @rtype: (dict, dict)
+ @return: (new_parameters, filled_parameters)
+
+ """
+ params_copy = copy.deepcopy(old_params)
+ for key, val in update_dict.iteritems():
+ if val == constants.VALUE_DEFAULT:
+ try:
+ del params_copy[key]
+ except KeyError:
+ pass
+ else:
+ params_copy[key] = val
+ utils.ForceDictType(params_copy, parameter_types)
+ params_filled = objects.FillDict(default_values, params_copy)
+ return (params_copy, params_filled)
+
def CheckPrereq(self):
"""Check prerequisites.
This only checks the instance list against the existing names.
"""
- force = self.force = self.op.force
+ self.force = self.op.force
# checking the new params on the primary/secondary nodes
instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
+ cluster = self.cluster = self.cfg.GetClusterInfo()
assert self.instance is not None, \
"Cannot retrieve locked instance %s" % self.op.instance_name
pnode = instance.primary_node
# hvparams processing
if self.op.hvparams:
- i_hvdict = copy.deepcopy(instance.hvparams)
- for key, val in self.op.hvparams.iteritems():
- if val == constants.VALUE_DEFAULT:
- try:
- del i_hvdict[key]
- except KeyError:
- pass
- elif val == constants.VALUE_NONE:
- i_hvdict[key] = None
- else:
- i_hvdict[key] = val
- cluster = self.cfg.GetClusterInfo()
- hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
- i_hvdict)
+ i_hvdict, hv_new = self._GetUpdatedParams(
+ instance.hvparams, self.op.hvparams,
+ cluster.hvparams[instance.hypervisor],
+ constants.HVS_PARAMETER_TYPES)
# local check
hypervisor.GetHypervisor(
instance.hypervisor).CheckParameterSyntax(hv_new)
# beparams processing
if self.op.beparams:
- i_bedict = copy.deepcopy(instance.beparams)
- for key, val in self.op.beparams.iteritems():
- if val == constants.VALUE_DEFAULT:
- try:
- del i_bedict[key]
- except KeyError:
- pass
- else:
- i_bedict[key] = val
- cluster = self.cfg.GetClusterInfo()
- be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
- i_bedict)
+ i_bedict, be_new = self._GetUpdatedParams(
+ instance.beparams, self.op.beparams,
+ cluster.beparams[constants.PP_DEFAULT],
+ constants.BES_PARAMETER_TYPES)
self.be_new = be_new # the new actual values
self.be_inst = i_bedict # the new dict (without defaults)
else:
instance.hypervisor)
nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
instance.hypervisor)
- if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
+ pninfo = nodeinfo[pnode]
+ msg = pninfo.fail_msg
+ if msg:
# Assume the primary node is unreachable and go ahead
- self.warn.append("Can't get info from primary node %s" % pnode)
+ self.warn.append("Can't get info from primary node %s: %s" %
+ (pnode, msg))
+ elif not isinstance(pninfo.payload.get('memory_free', None), int):
+ self.warn.append("Node data from primary node %s doesn't contain"
+ " free memory information" % pnode)
+ elif instance_info.fail_msg:
+ self.warn.append("Can't get instance runtime information: %s" %
+ instance_info.fail_msg)
else:
- if not instance_info.failed and instance_info.data:
- current_mem = instance_info.data['memory']
+ if instance_info.payload:
+ current_mem = int(instance_info.payload['memory'])
else:
# Assume instance not running
# (there is a slight race condition here, but it's not very probable,
# and we have no other way to check)
current_mem = 0
miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
- nodeinfo[pnode].data['memory_free'])
+ pninfo.payload['memory_free'])
if miss_mem > 0:
raise errors.OpPrereqError("This change will prevent the instance"
" from starting, due to %d MB of memory"
" missing on its primary node" % miss_mem)
if be_new[constants.BE_AUTO_BALANCE]:
- for node, nres in nodeinfo.iteritems():
+ for node, nres in nodeinfo.items():
if node not in instance.secondary_nodes:
continue
- if nres.failed or not isinstance(nres.data, dict):
- self.warn.append("Can't get info from secondary node %s" % node)
- elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
+ msg = nres.fail_msg
+ if msg:
+ self.warn.append("Can't get info from secondary node %s: %s" %
+ (node, msg))
+ elif not isinstance(nres.payload.get('memory_free', None), int):
+ self.warn.append("Secondary node %s didn't return free"
+ " memory information" % node)
+ elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
self.warn.append("Not enough memory to failover instance to"
" secondary node %s" % node)
# NIC processing
+ self.nic_pnew = {}
+ self.nic_pinst = {}
for nic_op, nic_dict in self.op.nics:
if nic_op == constants.DDM_REMOVE:
if not instance.nics:
raise errors.OpPrereqError("Invalid NIC index %s, valid values"
" are 0 to %d" %
(nic_op, len(instance.nics)))
- nic_bridge = nic_dict.get('bridge', None)
- if nic_bridge is not None:
- if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
- msg = ("Bridge '%s' doesn't exist on one of"
- " the instance nodes" % nic_bridge)
+ old_nic_params = instance.nics[nic_op].nicparams
+ old_nic_ip = instance.nics[nic_op].ip
+ else:
+ old_nic_params = {}
+ old_nic_ip = None
+
+ update_params_dict = dict([(key, nic_dict[key])
+ for key in constants.NICS_PARAMETERS
+ if key in nic_dict])
+
+ if 'bridge' in nic_dict:
+ update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
+
+ new_nic_params, new_filled_nic_params = \
+ self._GetUpdatedParams(old_nic_params, update_params_dict,
+ cluster.nicparams[constants.PP_DEFAULT],
+ constants.NICS_PARAMETER_TYPES)
+ objects.NIC.CheckParameterSyntax(new_filled_nic_params)
+ self.nic_pinst[nic_op] = new_nic_params
+ self.nic_pnew[nic_op] = new_filled_nic_params
+ new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
+
+ if new_nic_mode == constants.NIC_MODE_BRIDGED:
+ nic_bridge = new_filled_nic_params[constants.NIC_LINK]
+ msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
+ if msg:
+ msg = "Error checking bridges on node %s: %s" % (pnode, msg)
if self.force:
self.warn.append(msg)
else:
raise errors.OpPrereqError(msg)
+ if new_nic_mode == constants.NIC_MODE_ROUTED:
+ if 'ip' in nic_dict:
+ nic_ip = nic_dict['ip']
+ else:
+ nic_ip = old_nic_ip
+ if nic_ip is None:
+ raise errors.OpPrereqError('Cannot set the nic ip to None'
+ ' on a routed nic')
+ if 'mac' in nic_dict:
+ nic_mac = nic_dict['mac']
+ if nic_mac is None:
+ raise errors.OpPrereqError('Cannot set the nic mac to None')
+ elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
+ # otherwise generate the mac
+ nic_dict['mac'] = self.cfg.GenerateMAC()
+ else:
+ # or validate/reserve the current one
+ if self.cfg.IsMacInUse(nic_mac):
+ raise errors.OpPrereqError("MAC address %s already in use"
+ " in cluster" % nic_mac)
# DISK processing
if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
" an instance")
ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
ins_l = ins_l[pnode]
- if ins_l.failed or not isinstance(ins_l.data, list):
- raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
- if instance.name in ins_l.data:
+ msg = ins_l.fail_msg
+ if msg:
+ raise errors.OpPrereqError("Can't contact node %s: %s" %
+ (pnode, msg))
+ if instance.name in ins_l.payload:
raise errors.OpPrereqError("Instance is running, can't remove"
" disks.")
result = []
instance = self.instance
+ cluster = self.cluster
# disk changes
for disk_op, disk_dict in self.op.disks:
if disk_op == constants.DDM_REMOVE:
device_idx = len(instance.disks)
for node, disk in device.ComputeNodeTree(instance.primary_node):
self.cfg.SetDiskID(disk, node)
- rpc_result = self.rpc.call_blockdev_remove(node, disk)
- if rpc_result.failed or not rpc_result.data:
- self.proc.LogWarning("Could not remove disk/%d on node %s,"
- " continuing anyway", device_idx, node)
+ msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
+ if msg:
+ self.LogWarning("Could not remove disk/%d on node %s: %s,"
+ " continuing anyway", device_idx, node, msg)
result.append(("disk/%d" % device_idx, "remove"))
elif disk_op == constants.DDM_ADD:
# add a new disk
disk_idx_base = len(instance.disks)
new_disk = _GenerateDiskTemplate(self,
instance.disk_template,
- instance, instance.primary_node,
+ instance.name, instance.primary_node,
instance.secondary_nodes,
[disk_dict],
file_path,
file_driver,
disk_idx_base)[0]
- new_disk.mode = disk_dict['mode']
instance.disks.append(new_disk)
info = _GetInstanceInfoText(instance)
del instance.nics[-1]
result.append(("nic.%d" % len(instance.nics), "remove"))
elif nic_op == constants.DDM_ADD:
- # add a new nic
- if 'mac' not in nic_dict:
- mac = constants.VALUE_GENERATE
- else:
- mac = nic_dict['mac']
- if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
- mac = self.cfg.GenerateMAC()
- new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
- bridge=nic_dict.get('bridge', None))
+ # mac and bridge should be set, by now
+ mac = nic_dict['mac']
+ ip = nic_dict.get('ip', None)
+ nicparams = self.nic_pinst[constants.DDM_ADD]
+ new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
instance.nics.append(new_nic)
result.append(("nic.%d" % (len(instance.nics) - 1),
- "add:mac=%s,ip=%s,bridge=%s" %
- (new_nic.mac, new_nic.ip, new_nic.bridge)))
+ "add:mac=%s,ip=%s,mode=%s,link=%s" %
+ (new_nic.mac, new_nic.ip,
+ self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
+ self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
+ )))
else:
- # change a given nic
- for key in 'mac', 'ip', 'bridge':
+ for key in 'mac', 'ip':
if key in nic_dict:
setattr(instance.nics[nic_op], key, nic_dict[key])
- result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
+ if nic_op in self.nic_pnew:
+ instance.nics[nic_op].nicparams = self.nic_pnew[nic_op]
+ for key, val in nic_dict.iteritems():
+ result.append(("nic.%s/%d" % (key, nic_op), val))
# hvparams changes
if self.op.hvparams:
- instance.hvparams = self.hv_new
+ instance.hvparams = self.hv_inst
for key, val in self.op.hvparams.iteritems():
result.append(("hv/%s" % key, val))
rpcresult = self.rpc.call_export_list(self.nodes)
result = {}
for node in rpcresult:
- if rpcresult[node].failed:
+ if rpcresult[node].fail_msg:
result[node] = False
else:
- result[node] = rpcresult[node].data
+ result[node] = rpcresult[node].payload
return result
# remove it from its current node. In the future we could fix this by:
# - making a tasklet to search (share-lock all), then create the new one,
# then one to remove, after
- # - removing the removal operation altoghether
+ # - removing the removal operation altogether
self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
def DeclareLocks(self, level):
# This is wrong node name, not a non-locked node
raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
_CheckNodeOnline(self, self.dst_node.name)
+ _CheckNodeNotDrained(self, self.dst_node.name)
# instance disk type verification
for disk in self.instance.disks:
if self.op.shutdown:
# shutdown the instance, but not the disks
result = self.rpc.call_instance_shutdown(src_node, instance)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Could not shutdown instance %s on node %s" %
- (instance.name, src_node))
+ result.Raise("Could not shutdown instance %s on"
+ " node %s" % (instance.name, src_node))
vgname = self.cfg.GetVGName()
self.cfg.SetDiskID(disk, src_node)
try:
- for disk in instance.disks:
- # new_dev_name will be a snapshot of an lvm leaf of the one we passed
- new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
- if new_dev_name.failed or not new_dev_name.data:
- self.LogWarning("Could not snapshot block device %s on node %s",
- disk.logical_id[1], src_node)
+ for idx, disk in enumerate(instance.disks):
+ # result.payload will be a snapshot of an lvm leaf of the one we passed
+ result = self.rpc.call_blockdev_snapshot(src_node, disk)
+ msg = result.fail_msg
+ if msg:
+ self.LogWarning("Could not snapshot disk/%s on node %s: %s",
+ idx, src_node, msg)
snap_disks.append(False)
else:
+ disk_id = (vgname, result.payload)
new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
- logical_id=(vgname, new_dev_name.data),
- physical_id=(vgname, new_dev_name.data),
+ logical_id=disk_id, physical_id=disk_id,
iv_name=disk.iv_name)
snap_disks.append(new_dev)
finally:
- if self.op.shutdown and instance.status == "up":
- result = self.rpc.call_instance_start(src_node, instance, None)
- msg = result.RemoteFailMsg()
+ if self.op.shutdown and instance.admin_up:
+ result = self.rpc.call_instance_start(src_node, instance, None, None)
+ msg = result.fail_msg
if msg:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Could not start instance: %s" % msg)
if dev:
result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
instance, cluster_name, idx)
- if result.failed or not result.data:
- self.LogWarning("Could not export block device %s from node %s to"
- " node %s", dev.logical_id[1], src_node,
- dst_node.name)
- result = self.rpc.call_blockdev_remove(src_node, dev)
- if result.failed or not result.data:
- self.LogWarning("Could not remove snapshot block device %s from node"
- " %s", dev.logical_id[1], src_node)
+ msg = result.fail_msg
+ if msg:
+ self.LogWarning("Could not export disk/%s from node %s to"
+ " node %s: %s", idx, src_node, dst_node.name, msg)
+ msg = self.rpc.call_blockdev_remove(src_node, dev).fail_msg
+ if msg:
+ self.LogWarning("Could not remove snapshot for disk/%d from node"
+ " %s: %s", idx, src_node, msg)
result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
- if result.failed or not result.data:
- self.LogWarning("Could not finalize export for instance %s on node %s",
- instance.name, dst_node.name)
+ msg = result.fail_msg
+ if msg:
+ self.LogWarning("Could not finalize export for instance %s"
+ " on node %s: %s", instance.name, dst_node.name, msg)
nodelist = self.cfg.GetNodeList()
nodelist.remove(dst_node.name)
# on one-node clusters nodelist will be empty after the removal
# if we proceed the backup would be removed because OpQueryExports
# substitutes an empty list with the full cluster node list.
+ iname = instance.name
if nodelist:
exportlist = self.rpc.call_export_list(nodelist)
for node in exportlist:
- if exportlist[node].failed:
+ if exportlist[node].fail_msg:
continue
- if instance.name in exportlist[node].data:
- if not self.rpc.call_export_remove(node, instance.name):
+ if iname in exportlist[node].payload:
+ msg = self.rpc.call_export_remove(node, iname).fail_msg
+ if msg:
self.LogWarning("Could not remove older export for instance %s"
- " on node %s", instance.name, node)
+ " on node %s: %s", iname, node, msg)
class LURemoveExport(NoHooksLU):
fqdn_warn = True
instance_name = self.op.instance_name
- exportlist = self.rpc.call_export_list(self.acquired_locks[
- locking.LEVEL_NODE])
+ locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
+ exportlist = self.rpc.call_export_list(locked_nodes)
found = False
for node in exportlist:
- if exportlist[node].failed:
- self.LogWarning("Failed to query node %s, continuing" % node)
+ msg = exportlist[node].fail_msg
+ if msg:
+ self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
continue
- if instance_name in exportlist[node].data:
+ if instance_name in exportlist[node].payload:
found = True
result = self.rpc.call_export_remove(node, instance_name)
- if result.failed or not result.data:
+ msg = result.fail_msg
+ if msg:
logging.error("Could not remove export for instance %s"
- " on node %s", instance_name, node)
+ " on node %s: %s", instance_name, node, msg)
if fqdn_warn and not found:
feedback_fn("Export not found. If trying to remove an export belonging"
raise errors.OpExecError("Error during master delay test")
if self.op.on_nodes:
result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
- if not result:
- raise errors.OpExecError("Complete failure from rpc call")
for node, node_result in result.items():
- node_result.Raise()
- if not node_result.data:
- raise errors.OpExecError("Failure during rpc call to node %s,"
- " result: %s" % (node, node_result.data))
+ node_result.Raise("Failure during rpc call to node %s" % node)
class IAllocator(object):
"relocate_from",
]
- def __init__(self, lu, mode, name, **kwargs):
- self.lu = lu
+ def __init__(self, cfg, rpc, mode, name, **kwargs):
+ self.cfg = cfg
+ self.rpc = rpc
# init buffer variables
self.in_text = self.out_text = self.in_data = self.out_data = None
# init all input fields so that pylint is happy
This is the data that is independent of the actual operation.
"""
- cfg = self.lu.cfg
+ cfg = self.cfg
cluster_info = cfg.GetClusterInfo()
# cluster data
data = {
- "version": 1,
+ "version": constants.IALLOCATOR_VERSION,
"cluster_name": cfg.GetClusterName(),
"cluster_tags": list(cluster_info.GetTags()),
- "enable_hypervisors": list(cluster_info.enabled_hypervisors),
+ "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
# we don't have job IDs
}
iinfo = cfg.GetAllInstancesInfo().values()
elif self.mode == constants.IALLOCATOR_MODE_RELOC:
hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
- node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
- hypervisor_name)
- node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
- cluster_info.enabled_hypervisors)
- for nname in node_list:
+ node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
+ hypervisor_name)
+ node_iinfo = \
+ self.rpc.call_all_instances_info(node_list,
+ cluster_info.enabled_hypervisors)
+ for nname, nresult in node_data.items():
+ # first fill in static (config-based) values
ninfo = cfg.GetNodeInfo(nname)
- node_data[nname].Raise()
- if not isinstance(node_data[nname].data, dict):
- raise errors.OpExecError("Can't get data for node %s" % nname)
- remote_info = node_data[nname].data
- for attr in ['memory_total', 'memory_free', 'memory_dom0',
- 'vg_size', 'vg_free', 'cpu_total']:
- if attr not in remote_info:
- raise errors.OpExecError("Node '%s' didn't return attribute '%s'" %
- (nname, attr))
- try:
- remote_info[attr] = int(remote_info[attr])
- except ValueError, err:
- raise errors.OpExecError("Node '%s' returned invalid value for '%s':"
- " %s" % (nname, attr, str(err)))
- # compute memory used by primary instances
- i_p_mem = i_p_up_mem = 0
- for iinfo, beinfo in i_list:
- if iinfo.primary_node == nname:
- i_p_mem += beinfo[constants.BE_MEMORY]
- if iinfo.name not in node_iinfo[nname]:
- i_used_mem = 0
- else:
- i_used_mem = int(node_iinfo[nname][iinfo.name]['memory'])
- i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
- remote_info['memory_free'] -= max(0, i_mem_diff)
-
- if iinfo.status == "up":
- i_p_up_mem += beinfo[constants.BE_MEMORY]
-
- # compute memory used by instances
pnr = {
"tags": list(ninfo.GetTags()),
- "total_memory": remote_info['memory_total'],
- "reserved_memory": remote_info['memory_dom0'],
- "free_memory": remote_info['memory_free'],
- "i_pri_memory": i_p_mem,
- "i_pri_up_memory": i_p_up_mem,
- "total_disk": remote_info['vg_size'],
- "free_disk": remote_info['vg_free'],
"primary_ip": ninfo.primary_ip,
"secondary_ip": ninfo.secondary_ip,
- "total_cpus": remote_info['cpu_total'],
"offline": ninfo.offline,
+ "drained": ninfo.drained,
+ "master_candidate": ninfo.master_candidate,
}
+
+ if not ninfo.offline:
+ nresult.Raise("Can't get data for node %s" % nname)
+ node_iinfo[nname].Raise("Can't get node instance info from node %s" %
+ nname)
+ remote_info = nresult.payload
+ for attr in ['memory_total', 'memory_free', 'memory_dom0',
+ 'vg_size', 'vg_free', 'cpu_total']:
+ if attr not in remote_info:
+ raise errors.OpExecError("Node '%s' didn't return attribute"
+ " '%s'" % (nname, attr))
+ if not isinstance(remote_info[attr], int):
+ raise errors.OpExecError("Node '%s' returned invalid value"
+ " for '%s': %s" %
+ (nname, attr, remote_info[attr]))
+ # compute memory used by primary instances
+ i_p_mem = i_p_up_mem = 0
+ for iinfo, beinfo in i_list:
+ if iinfo.primary_node == nname:
+ i_p_mem += beinfo[constants.BE_MEMORY]
+ if iinfo.name not in node_iinfo[nname].payload:
+ i_used_mem = 0
+ else:
+ i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
+ i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
+ remote_info['memory_free'] -= max(0, i_mem_diff)
+
+ if iinfo.admin_up:
+ i_p_up_mem += beinfo[constants.BE_MEMORY]
+
+ # compute memory used by instances
+ pnr_dyn = {
+ "total_memory": remote_info['memory_total'],
+ "reserved_memory": remote_info['memory_dom0'],
+ "free_memory": remote_info['memory_free'],
+ "total_disk": remote_info['vg_size'],
+ "free_disk": remote_info['vg_free'],
+ "total_cpus": remote_info['cpu_total'],
+ "i_pri_memory": i_p_mem,
+ "i_pri_up_memory": i_p_up_mem,
+ }
+ pnr.update(pnr_dyn)
+
node_results[nname] = pnr
data["nodes"] = node_results
# instance data
instance_data = {}
for iinfo, beinfo in i_list:
- nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
- for n in iinfo.nics]
+ nic_data = []
+ for nic in iinfo.nics:
+ filled_params = objects.FillDict(
+ cluster_info.nicparams[constants.PP_DEFAULT],
+ nic.nicparams)
+ nic_dict = {"mac": nic.mac,
+ "ip": nic.ip,
+ "mode": filled_params[constants.NIC_MODE],
+ "link": filled_params[constants.NIC_LINK],
+ }
+ if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
+ nic_dict["bridge"] = filled_params[constants.NIC_LINK]
+ nic_data.append(nic_dict)
pir = {
"tags": list(iinfo.GetTags()),
- "should_run": iinfo.status == "up",
+ "admin_up": iinfo.admin_up,
"vcpus": beinfo[constants.BE_VCPUS],
"memory": beinfo[constants.BE_MEMORY],
"os": iinfo.os,
- "nodes": list(iinfo.all_nodes),
+ "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
"nics": nic_data,
- "disks": [{"size": dsk.size, "mode": "w"} for dsk in iinfo.disks],
+ "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
"disk_template": iinfo.disk_template,
"hypervisor": iinfo.hypervisor,
}
+ pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
+ pir["disks"])
instance_data[iinfo.name] = pir
data["instances"] = instance_data
"""
data = self.in_data
- if len(self.disks) != 2:
- raise errors.OpExecError("Only two-disk configurations supported")
disk_space = _ComputeDiskSize(self.disk_template, self.disks)
done.
"""
- instance = self.lu.cfg.GetInstanceInfo(self.name)
+ instance = self.cfg.GetInstanceInfo(self.name)
if instance is None:
raise errors.ProgrammerError("Unknown instance '%s' passed to"
" IAllocator" % self.name)
"""
if call_fn is None:
- call_fn = self.lu.rpc.call_iallocator_runner
- data = self.in_text
-
- result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
- result.Raise()
-
- if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
- raise errors.OpExecError("Invalid result from master iallocator runner")
+ call_fn = self.rpc.call_iallocator_runner
- rcode, stdout, stderr, fail = result.data
+ result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
+ result.Raise("Failure while running the iallocator script")
- if rcode == constants.IARUN_NOTFOUND:
- raise errors.OpExecError("Can't find allocator '%s'" % name)
- elif rcode == constants.IARUN_FAILURE:
- raise errors.OpExecError("Instance allocator call failed: %s,"
- " output: %s" % (fail, stdout+stderr))
- self.out_text = stdout
+ self.out_text = result.payload
if validate:
self._ValidateResult()
" 'nics' parameter")
if not isinstance(self.op.disks, list):
raise errors.OpPrereqError("Invalid parameter 'disks'")
- if len(self.op.disks) != 2:
- raise errors.OpPrereqError("Only two-disk configurations supported")
for row in self.op.disks:
if (not isinstance(row, dict) or
"size" not in row or
row["mode"] not in ['r', 'w']):
raise errors.OpPrereqError("Invalid contents of the"
" 'disks' parameter")
- if self.op.hypervisor is None:
+ if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
self.op.hypervisor = self.cfg.GetHypervisorType()
elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
if not hasattr(self.op, "name"):
"""
if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
- ial = IAllocator(self,
+ ial = IAllocator(self.cfg, self.rpc,
mode=self.op.mode,
name=self.op.name,
mem_size=self.op.mem_size,
hypervisor=self.op.hypervisor,
)
else:
- ial = IAllocator(self,
+ ial = IAllocator(self.cfg, self.rpc,
mode=self.op.mode,
name=self.op.name,
relocate_from=list(self.relocate_from),