import os
import os.path
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
Subclasses must follow these rules:
- implement ExpandNames
- - implement CheckPrereq
- - implement Exec
+ - implement CheckPrereq (except when tasklets are used)
+ - implement Exec (except when tasklets are used)
- implement BuildHooksEnv
- redefine HPATH and HTYPE
- optionally redefine their run requirements:
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
+
+ # Tasklets
+ self.tasklets = None
for attr_name in self._OP_REQP:
attr_val = getattr(op, attr_name, None)
if attr_val is None:
raise errors.OpPrereqError("Required parameter '%s' missing" %
attr_name)
+
self.CheckArguments()
def __GetSSH(self):
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
level you can modify self.share_locks, setting a true value (usually 1) for
that level. By default locks are not shared.
+ This function can also define a list of tasklets, which then will be
+ executed in order instead of the usual LU-level CheckPrereq and Exec
+ functions, if those are not defined by the LU.
+
Examples::
# Acquire all nodes and one instance
their canonical form if it hasn't been done by ExpandNames before.
"""
- raise NotImplementedError
+ if self.tasklets is not None:
+ for (idx, tl) in enumerate(self.tasklets):
+ logging.debug("Checking prerequisites for tasklet %s/%s",
+ idx + 1, len(self.tasklets))
+ tl.CheckPrereq()
+ else:
+ raise NotImplementedError
def Exec(self, feedback_fn):
"""Execute the LU.
code, or expected.
"""
- raise NotImplementedError
+ if self.tasklets is not None:
+ for (idx, tl) in enumerate(self.tasklets):
+ logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
+ tl.Exec(feedback_fn)
+ else:
+ raise NotImplementedError
def BuildHooksEnv(self):
"""Build hooks environment for this LU.
HTYPE = None
+class Tasklet:
+ """Tasklet base class.
+
+ Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
+ they can mix legacy code with tasklets. Locking needs to be done in the LU,
+ tasklets know nothing about locks.
+
+ Subclasses must follow these rules:
+ - Implement CheckPrereq
+ - Implement Exec
+
+ """
+ def __init__(self, lu):
+ self.lu = lu
+
+ # Shortcuts
+ self.cfg = lu.cfg
+ self.rpc = lu.rpc
+
+ def CheckPrereq(self):
+ """Check prerequisites for this tasklets.
+
+ This method should check whether the prerequisites for the execution of
+ this tasklet are fulfilled. It can do internode communication, but it
+ should be idempotent - no cluster or system changes are allowed.
+
+ The method should raise errors.OpPrereqError in case something is not
+ fulfilled. Its return value is ignored.
+
+ This method should also update all parameters to their canonical form if it
+ hasn't been done before.
+
+ """
+ raise NotImplementedError
+
+ def Exec(self, feedback_fn):
+ """Execute the tasklet.
+
+ This method should implement the actual work. It should raise
+ errors.OpExecError for failures that are somewhat dealt with in code, or
+ expected.
+
+ """
+ raise NotImplementedError
+
+
def _GetWantedNodes(lu, nodes):
"""Returns list of checked and expanded node names.
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
- memory, vcpus, nics, disk_template, disks):
+ 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.
@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 distk template of the instance
+ @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
"INSTANCE_MEMORY": memory,
"INSTANCE_VCPUS": vcpus,
"INSTANCE_DISK_TEMPLATE": disk_template,
+ "INSTANCE_HYPERVISOR": hypervisor_name,
}
if nics:
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 _PreBuildNICHooksList(lu, nics):
+def _NICListToTuple(lu, nics):
"""Build a list of nic information tuples.
- This list is suitable to be passed to _BuildInstanceHookEnv.
+ 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
@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,
'status': instance.admin_up,
'memory': bep[constants.BE_MEMORY],
'vcpus': bep[constants.BE_VCPUS],
- 'nics': _PreBuildNICHooksList(lu, 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)
if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
if brlist:
result = lu.rpc.call_bridges_exist(target_node, brlist)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Error checking bridges on destination node"
- " '%s': %s" % (target_node, msg))
+ result.Raise("Error checking bridges on destination node '%s'" %
+ target_node, prereq=True)
def _CheckInstanceBridgesExist(lu, instance, node=None):
"""
if node is None:
- node=instance.primary_node
+ node = instance.primary_node
_CheckNicsBridgesExist(lu, instance.nics, node)
+def _GetNodePrimaryInstances(cfg, node_name):
+ """Returns primary instances on a node.
+
+ """
+ instances = []
+
+ for (_, inst) in cfg.GetAllInstancesInfo().iteritems():
+ if node_name == inst.primary_node:
+ instances.append(inst)
+
+ return instances
+
+
+def _GetNodeSecondaryInstances(cfg, node_name):
+ """Returns secondary instances on a node.
+
+ """
+ instances = []
+
+ for (_, inst) in cfg.GetAllInstancesInfo().iteritems():
+ if node_name in inst.secondary_nodes:
+ instances.append(inst)
+
+ return instances
+
+
class LUDestroyCluster(NoHooksLU):
"""Logical unit for destroying the cluster.
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,
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
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 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.
"""
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 = "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:
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.
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
node_res = node_lvs[node]
if node_res.offline:
continue
- msg = node_res.RemoteFailMsg()
+ msg = node_res.fail_msg
if msg:
logging.warning("Error enumerating LVs on node %s: %s", node, msg)
res_nodes[node] = msg
# 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():
- msg = to_result.RemoteFailMsg()
- if msg:
- msg = ("Copy of file %s to node %s failed: %s" %
- (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
- self.proc.LogWarning(msg)
+ 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
"""
if self.op.vg_name:
vglist = self.rpc.call_vg_list(node_list)
for node in node_list:
- msg = vglist[node].RemoteFailMsg()
+ msg = vglist[node].fail_msg
if msg:
# ignoring down node
self.LogWarning("Error while gathering data on node %s"
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
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.
constants.SSH_KNOWN_HOSTS_FILE,
constants.RAPI_CERT_FILE,
constants.RAPI_USERS_FILE,
+ constants.HMAC_CLUSTER_KEY,
])
enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
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.RemoteFailMsg()
- if msg:
- msg = ("Copy of file %s to node %s failed: %s" %
- (fname, to_node, msg))
- lu.proc.LogWarning(msg)
+ 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):
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)
- msg = rstats.RemoteFailMsg()
+ msg = rstats.fail_msg
if msg:
lu.LogWarning("Can't get any data from node %s: %s", node, msg)
retries += 1
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)
- msg = rstats.RemoteFailMsg()
+ msg = rstats.fail_msg
if msg:
lu.LogWarning("Can't find disk on node %s: %s", node, msg)
result = False
@rtype: dict
@return: a dictionary with osnames as keys and as value another map, with
- nodes as keys and list of OS objects as values, eg::
+ 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, "")]}
}
"""
# 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].failed]
- for node_name, nr in rlist.iteritems():
- if nr.failed or not nr.data:
+ 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] = {}
+ all_os[name] = {}
for nname in good_nodes:
- all_os[os_obj.name][nname] = []
- all_os[os_obj.name][node_name].append(os_obj)
+ all_os[name][nname] = []
+ all_os[name][node_name].append((path, status, diagnose))
return all_os
def Exec(self, feedback_fn):
"""
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)
"master",
"offline",
"drained",
+ "role",
)
def ExpandNames(self):
self.cfg.GetHypervisorType())
for name in nodenames:
nodeinfo = node_data[name]
- if not nodeinfo.RemoteFailMsg() and nodeinfo.payload:
+ if not nodeinfo.fail_msg and nodeinfo.payload:
nodeinfo = nodeinfo.payload
fn = utils.TryConvert
live_data[name] = {
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:
return output
+class LUQueryNodeStorage(NoHooksLU):
+ """Logical unit for getting information on storage units on node(s).
+
+ """
+ _OP_REQP = ["nodes", "storage_type", "output_fields"]
+ REQ_BGL = False
+ _FIELDS_STATIC = utils.FieldSet("node")
+
+ def ExpandNames(self):
+ storage_type = self.op.storage_type
+
+ if storage_type not in constants.VALID_STORAGE_FIELDS:
+ raise errors.OpPrereqError("Unknown storage type: %s" % storage_type)
+
+ dynamic_fields = constants.VALID_STORAGE_FIELDS[storage_type]
+
+ _CheckOutputFields(static=self._FIELDS_STATIC,
+ dynamic=utils.FieldSet(*dynamic_fields),
+ selected=self.op.output_fields)
+
+ self.needed_locks = {}
+ self.share_locks[locking.LEVEL_NODE] = 1
+
+ if self.op.nodes:
+ self.needed_locks[locking.LEVEL_NODE] = \
+ _GetWantedNodes(self, self.op.nodes)
+ else:
+ self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This checks that the fields required are valid output fields.
+
+ """
+ self.op.name = getattr(self.op, "name", None)
+
+ self.nodes = self.acquired_locks[locking.LEVEL_NODE]
+
+ def Exec(self, feedback_fn):
+ """Computes the list of nodes and their attributes.
+
+ """
+ # Special case for file storage
+ if self.op.storage_type == constants.ST_FILE:
+ st_args = [self.cfg.GetFileStorageDir()]
+ else:
+ st_args = []
+
+ # Always get name to sort by
+ if constants.SF_NAME in self.op.output_fields:
+ fields = self.op.output_fields[:]
+ else:
+ fields = [constants.SF_NAME] + self.op.output_fields
+
+ # Never ask for node as it's only known to the LU
+ while "node" in fields:
+ fields.remove("node")
+
+ field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
+ name_idx = field_idx[constants.SF_NAME]
+
+ data = self.rpc.call_storage_list(self.nodes,
+ self.op.storage_type, st_args,
+ self.op.name, fields)
+
+ result = []
+
+ for node in utils.NiceSort(self.nodes):
+ nresult = data[node]
+ if nresult.offline:
+ continue
+
+ msg = nresult.fail_msg
+ if msg:
+ self.LogWarning("Can't get storage data from node %s: %s", node, msg)
+ continue
+
+ rows = dict([(row[name_idx], row) for row in nresult.payload])
+
+ for name in utils.NiceSort(rows.keys()):
+ row = rows[name]
+
+ out = []
+
+ for field in self.op.output_fields:
+ if field == "node":
+ val = node
+ elif field in field_idx:
+ val = row[field_idx[field]]
+ else:
+ raise errors.ParameterError(field)
+
+ out.append(val)
+
+ result.append(out)
+
+ return result
+
+
class LUAddNode(LogicalUnit):
"""Logical unit for adding node to the cluster.
- 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, drained=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])
-
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Cannot transfer ssh keys to the"
- " new node: %s" % msg)
+ result.Raise("Cannot transfer ssh keys to the new node")
# Add node to our /etc/hosts, and add key to known_hosts
if self.cfg.GetClusterInfo().modify_etc_hosts:
if new_node.secondary_ip != new_node.primary_ip:
result = self.rpc.call_node_has_ip_address(new_node.name,
new_node.secondary_ip)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Failure checking secondary ip"
- " on node %s: %s" % (new_node.name, msg))
+ 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"
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].data['nodelist'][failed]))
+ (verifier, nl_payload[failed]))
raise errors.OpExecError("ssh/hostname verification failed.")
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)
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)
- msg = rrc.RemoteFailMsg()
+ msg = rrc.fail_msg
if msg:
self.LogWarning("Node failed to demote itself: %s" % msg)
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"))
"""
result = self.rpc.call_node_powercycle(self.op.node_name,
self.cfg.GetHypervisorType())
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Failed to schedule the reboot: %s" % msg)
+ result.Raise("Failed to schedule the reboot")
return result.payload
"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": dict([(hypervisor, cluster.hvparams[hypervisor])
- for hypervisor in cluster.enabled_hypervisors]),
+ "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,
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)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
lu.proc.LogWarning("Could not prepare block device %s on node %s"
" (is_primary=False, pass=1): %s",
continue
lu.cfg.SetDiskID(node_disk, node)
result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
lu.proc.LogWarning("Could not prepare block device %s on node %s"
" (is_primary=True, pass=2): %s",
"""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)
"""
pnode = instance.primary_node
- ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])
- ins_l = ins_l[pnode]
- msg = ins_l.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't contact node %s: %s" % (pnode, msg))
+ 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"
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)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
lu.LogWarning("Could not shutdown block device %s on node %s: %s",
disk.iv_name, node, msg)
"""
nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
- msg = nodeinfo[node].RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Can't get data from node %s: %s" % (node, msg))
+ 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"
_CheckNodeOnline(self, instance.primary_node)
bep = self.cfg.GetClusterInfo().FillBE(instance)
- # check bridges existance
+ # check bridges existence
_CheckInstanceBridgesExist(self, instance)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- msg = remote_info.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Error checking node %s: %s" %
- (instance.primary_node, msg))
+ 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,
result = self.rpc.call_instance_start(node_current, instance,
self.hvparams, self.beparams)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Could not start instance: %s" % msg)
_CheckNodeOnline(self, instance.primary_node)
- # check bridges existance
+ # check bridges existence
_CheckInstanceBridgesExist(self, instance)
def Exec(self, feedback_fn):
self.cfg.SetDiskID(disk, node_current)
result = self.rpc.call_instance_reboot(node_current, instance,
reboot_type)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not reboot instance: %s" % msg)
+ result.Raise("Could not reboot instance")
else:
result = self.rpc.call_instance_shutdown(node_current, instance)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not shutdown instance for"
- " full reboot: %s" % msg)
+ 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, None, None)
- msg = result.RemoteFailMsg()
+ 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)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
self.proc.LogWarning("Could not shutdown instance: %s" % msg)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- msg = remote_info.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Error checking node %s: %s" %
- (instance.primary_node, msg))
+ 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,
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
try:
feedback_fn("Running the instance OS create scripts...")
result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
- 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.Raise("Could not install OS for instance %s on node %s" %
+ (inst.name, inst.primary_node))
finally:
_ShutdownInstanceDisks(self, inst)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- msg = remote_info.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Error checking node %s: %s" %
- (instance.primary_node, msg))
+ 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,
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)
- msg = result.RemoteFailMsg()
+ 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): %s" %
instance.name, instance.primary_node)
result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
if self.op.ignore_failures:
feedback_fn("Warning: can't shutdown instance: %s" % msg)
_FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
"admin_state",
"disk_template", "ip", "mac", "bridge",
+ "nic_mode", "nic_link",
"sda_size", "sdb_size", "vcpus", "tags",
"network_port", "beparams",
r"(disk)\.(size)/([0-9]+)",
r"(disk)\.(sizes)", "disk_usage",
- r"(nic)\.(mac|ip|bridge)/([0-9]+)",
- r"(nic)\.(macs|ips|bridges)",
+ 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
if result.offline:
# offline nodes will be in both lists
off_nodes.append(name)
- if result.failed or result.RemoteFailMsg():
+ if result.failed or result.fail_msg:
bad_nodes.append(name)
else:
if result.payload:
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":
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 = [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)
_CheckNodeNotDrained(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)
+ 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
_CheckInstanceBridgesExist(self, instance, node=target_node)
instance.name, source_node)
result = self.rpc.call_instance_shutdown(source_node, instance)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
if self.op.ignore_consistency:
self.proc.LogWarning("Could not shutdown instance %s on node %s."
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)
feedback_fn("* starting the instance on the target node")
result = self.rpc.call_instance_start(target_node, instance, None, None)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
_ShutdownInstanceDisks(self, instance)
raise errors.OpExecError("Could not start instance %s on node %s: %s" %
def ExpandNames(self):
self._ExpandAndLockInstance()
+
self.needed_locks[locking.LEVEL_NODE] = []
self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
+ self._migrater = TLMigrateInstance(self, self.op.instance_name,
+ self.op.live, self.op.cleanup)
+ self.tasklets = [self._migrater]
+
def DeclareLocks(self, level):
if level == locking.LEVEL_NODE:
self._LockInstancesNodes()
This runs on master, primary and secondary nodes of the instance.
"""
- env = _BuildInstanceHookEnvByObject(self, self.instance)
+ instance = self._migrater.instance
+ env = _BuildInstanceHookEnvByObject(self, instance)
env["MIGRATE_LIVE"] = self.op.live
env["MIGRATE_CLEANUP"] = self.op.cleanup
- nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
+ nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
return env, nl, nl
+
+class LUMigrateNode(LogicalUnit):
+ """Migrate all instances from a node.
+
+ """
+ HPATH = "node-migrate"
+ HTYPE = constants.HTYPE_NODE
+ _OP_REQP = ["node_name", "live"]
+ REQ_BGL = False
+
+ def ExpandNames(self):
+ self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
+ if self.op.node_name is None:
+ raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name)
+
+ self.needed_locks = {
+ locking.LEVEL_NODE: [self.op.node_name],
+ }
+
+ self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
+
+ # Create tasklets for migrating instances for all instances on this node
+ names = []
+ tasklets = []
+
+ for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
+ logging.debug("Migrating instance %s", inst.name)
+ names.append(inst.name)
+
+ tasklets.append(TLMigrateInstance(self, inst.name, self.op.live, False))
+
+ self.tasklets = tasklets
+
+ # Declare instance locks
+ self.needed_locks[locking.LEVEL_INSTANCE] = names
+
+ def DeclareLocks(self, level):
+ if level == locking.LEVEL_NODE:
+ self._LockInstancesNodes()
+
+ def BuildHooksEnv(self):
+ """Build hooks env.
+
+ This runs on the master, the primary and all the secondaries.
+
+ """
+ env = {
+ "NODE_NAME": self.op.node_name,
+ }
+
+ nl = [self.cfg.GetMasterNode()]
+
+ return (env, nl, nl)
+
+
+class TLMigrateInstance(Tasklet):
+ def __init__(self, lu, instance_name, live, cleanup):
+ """Initializes this class.
+
+ """
+ Tasklet.__init__(self, lu)
+
+ # Parameters
+ self.instance_name = instance_name
+ self.live = live
+ self.cleanup = cleanup
+
def CheckPrereq(self):
"""Check prerequisites.
"""
instance = self.cfg.GetInstanceInfo(
- self.cfg.ExpandInstanceName(self.op.instance_name))
+ self.cfg.ExpandInstanceName(self.instance_name))
if instance is None:
raise errors.OpPrereqError("Instance '%s' not known" %
- self.op.instance_name)
+ self.instance_name)
if instance.disk_template != constants.DT_DRBD8:
raise errors.OpPrereqError("Instance's disk layout is not"
# check bridge existance
_CheckInstanceBridgesExist(self, instance, node=target_node)
- if not self.op.cleanup:
+ if not self.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))
+ 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:
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():
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't contact node %s: %s" % (node, msg))
+ result.Raise("Can't contact node %s" % node)
runningon_source = instance.name in ins_l[source_node].payload
runningon_target = instance.name in ins_l[target_node].payload
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.lu.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,
migration_info,
False)
- abort_msg = abort_result.RemoteFailMsg()
+ abort_msg = abort_result.fail_msg
if abort_msg:
logging.error("Aborting migration failed on target node %s: %s" %
(target_node, abort_msg))
# First get the migration information from the remote node
result = self.rpc.call_migration_info(source_node, instance)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
log_err = ("Failed fetching source migration information from %s: %s" %
(source_node, msg))
migration_info,
self.nodes_ip[target_node])
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
logging.error("Instance pre-migration failed, trying to revert"
" disk status: %s", msg)
time.sleep(10)
result = self.rpc.call_instance_migrate(source_node, instance,
self.nodes_ip[target_node],
- self.op.live)
- msg = result.RemoteFailMsg()
+ self.live)
+ msg = result.fail_msg
if msg:
logging.error("Instance migration failed, trying to revert"
" disk status: %s", msg)
instance,
migration_info,
True)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
logging.error("Instance migration succeeded, but finalization failed:"
" %s" % msg)
"""Perform the migration.
"""
+ feedback_fn("Migrating instance %s" % self.instance.name)
+
self.feedback_fn = feedback_fn
self.source_node = self.instance.primary_node
self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
}
- if self.op.cleanup:
+
+ if self.cleanup:
return self._ExecCleanup()
else:
return self._ExecMigration()
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.payload
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
[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")
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
for device in instance.disks:
for node, disk in device.ComputeNodeTree(instance.primary_node):
lu.cfg.SetDiskID(disk, node)
- msg = lu.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
+ 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)
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)
+ 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 all_result
info = hvinfo[node]
if info.offline:
continue
- msg = info.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Hypervisor parameter validation"
- " failed on node %s: %s" % (node, msg))
+ info.Raise("Hypervisor parameter validation failed on node %s" % node)
class LUCreateInstance(LogicalUnit):
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.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
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")
+ 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:
"""
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,
os_type=self.op.os_type,
memory=self.be_full[constants.BE_MEMORY],
vcpus=self.be_full[constants.BE_VCPUS],
- nics=_PreBuildNICHooksList(self, 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] +
exp_list = self.rpc.call_export_list(locked_nodes)
found = False
for node in exp_list:
- if exp_list[node].RemoteFailMsg():
+ if exp_list[node].fail_msg:
continue
if src_path in exp_list[node].payload:
found = True
_CheckNodeOnline(self, src_node)
result = self.rpc.call_export_info(src_node, src_path)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("No export or invalid export found in"
- " dir %s: %s" % (src_path, msg))
+ result.Raise("No export or invalid export found in dir %s" % src_path)
export_info = objects.SerializableConfigParser.Loads(str(result.payload))
if not export_info.has_section(constants.INISECT_EXP):
self.op.hypervisor)
for node in nodenames:
info = nodeinfo[node]
- msg = info.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Cannot get current information"
- " from node %s: %s" % (node, msg))
+ 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):
# 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)
+ 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)
self.be_full[constants.BE_MEMORY],
self.op.hypervisor)
+ self.dry_run_result = list(nodenames)
+
def Exec(self, feedback_fn):
"""Create and add the instance to the cluster.
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, False)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not add os for instance %s"
- " on node %s: %s" %
- (instance, pnode_name, msg))
+ 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)
- msg = import_result.RemoteFailMsg()
+ 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))
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, None)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not start instance: %s" % msg)
+ 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]
- msg = node_insts.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't get node information from %s: %s" %
- (node, msg))
+ node_insts.Raise("Can't get node information from %s" % node)
if instance.name not in node_insts.payload:
raise errors.OpExecError("Instance %s is not running." % instance.name)
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")
+ TLReplaceDisks.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 = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
+ self.op.iallocator, self.op.remote_node,
+ self.op.disks)
+
+ self.tasklets = [self.replacer]
+
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.
+
+ This runs on the master, the primary and all the secondaries.
"""
- ial = IAllocator(self,
- mode=constants.IALLOCATOR_MODE_RELOC,
- name=self.op.instance_name,
- relocate_from=[self.sec_node])
+ instance = self.replacer.instance
+ env = {
+ "MODE": self.op.mode,
+ "NEW_SECONDARY": self.op.remote_node,
+ "OLD_SECONDARY": instance.secondary_nodes[0],
+ }
+ env.update(_BuildInstanceHookEnvByObject(self, instance))
+ nl = [
+ self.cfg.GetMasterNode(),
+ instance.primary_node,
+ ]
+ if self.op.remote_node is not None:
+ nl.append(self.op.remote_node)
+ return env, nl, nl
- 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)
+class LUEvacuateNode(LogicalUnit):
+ """Relocate the secondary instances from a node.
+
+ """
+ HPATH = "node-evacuate"
+ HTYPE = constants.HTYPE_NODE
+ _OP_REQP = ["node_name"]
+ REQ_BGL = False
+
+ def CheckArguments(self):
+ if not hasattr(self.op, "remote_node"):
+ self.op.remote_node = None
+ if not hasattr(self.op, "iallocator"):
+ self.op.iallocator = None
+
+ TLReplaceDisks.CheckArguments(constants.REPLACE_DISK_CHG,
+ self.op.remote_node,
+ self.op.iallocator)
+
+ def ExpandNames(self):
+ self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
+ if self.op.node_name is None:
+ raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name)
+
+ self.needed_locks = {}
+
+ # Declare node locks
+ 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:
+ raise errors.OpPrereqError("Invalid parameters")
+
+ # Create tasklets for replacing disks for all secondary instances on this
+ # node
+ names = []
+ tasklets = []
+
+ for inst in _GetNodeSecondaryInstances(self.cfg, self.op.node_name):
+ logging.debug("Replacing disks for instance %s", inst.name)
+ names.append(inst.name)
+
+ replacer = TLReplaceDisks(self, inst.name, constants.REPLACE_DISK_CHG,
+ self.op.iallocator, self.op.remote_node, [])
+ tasklets.append(replacer)
+
+ self.tasklets = tasklets
+ self.instance_names = names
+
+ # Declare instance locks
+ self.needed_locks[locking.LEVEL_INSTANCE] = self.instance_names
+
+ 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.
+ if (level == locking.LEVEL_NODE and
+ self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
+ self._LockInstancesNodes()
def BuildHooksEnv(self):
"""Build hooks env.
"""
env = {
- "MODE": self.op.mode,
- "NEW_SECONDARY": self.op.remote_node,
- "OLD_SECONDARY": self.instance.secondary_nodes[0],
+ "NODE_NAME": self.op.node_name,
}
- env.update(_BuildInstanceHookEnvByObject(self, self.instance))
- nl = [
- self.cfg.GetMasterNode(),
- self.instance.primary_node,
- ]
+
+ nl = [self.cfg.GetMasterNode()]
+
if self.op.remote_node is not None:
+ env["NEW_SECONDARY"] = self.op.remote_node
nl.append(self.op.remote_node)
- return env, nl, nl
+
+ return (env, nl, nl)
+
+
+class TLReplaceDisks(Tasklet):
+ """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.
+
+ """
+ Tasklet.__init__(self, lu)
+
+ # Parameters
+ self.instance_name = instance_name
+ self.mode = mode
+ self.iallocator_name = iallocator_name
+ self.remote_node = remote_node
+ self.disks = disks
+
+ # 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):
+ """Helper function for users of this class.
+
+ """
+ # 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.
"""
- 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.instance = self.cfg.GetInstanceInfo(self.instance_name)
+ assert self.instance is not None, \
+ "Cannot retrieve locked instance %s" % self.instance_name
- if instance.disk_template != constants.DT_DRBD8:
+ 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
- _CheckNodeNotDrained(self, remote_node)
- else:
- raise errors.ProgrammerError("Unhandled disk replace mode")
+ 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]
- _CheckNodeOnline(self, n1)
- _CheckNodeOnline(self, n2)
+ 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]
- if not self.op.disks:
- self.op.disks = range(len(instance.disks))
+ 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]
- for disk_idx in self.op.disks:
- instance.FindDisk(disk_idx)
+ _CheckNodeNotDrained(self.lu, remote_node)
- def _ExecD8DiskOnly(self, feedback_fn):
- """Replace a disk on the primary or secondary for dbrd8.
+ else:
+ raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
+ self.mode)
- The algorithm for replace is quite complicated:
+ for node in check_nodes:
+ _CheckNodeOnline(self.lu, node)
- 1. for each disk to be replaced:
+ # If not specified all disks should be replaced
+ if not self.disks:
+ self.disks = range(len(self.instance.disks))
- 1. create new LVs on the target node with unique names
- 1. detach old LVs from the drbd device
- 1. rename old LVs to name_replaced.<time_t>
- 1. rename new LVs to old LVs
- 1. attach the new LVs (with the old names now) to the drbd device
+ # Check whether disks are valid
+ for disk_idx in self.disks:
+ self.instance.FindDisk(disk_idx)
- 1. wait for sync across all devices
+ # Get secondary node IP addresses
+ node_2nd_ip = {}
- 1. for each modified disk:
+ 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
- 1. remove old LVs (which have the name name_replaces.<time_t>)
+ self.node_secondary_ip = node_2nd_ip
- Failures are not very well handled.
+ def Exec(self, feedback_fn):
+ """Execute disk replacement.
+
+ This dispatches the disk replacement to the appropriate handler.
"""
- steps_total = 6
- warning, info = (self.proc.LogWarning, self.proc.LogInfo)
- instance = self.instance
- iv_names = {}
+ feedback_fn("Replacing disks for %s" % self.instance.name)
+
+ 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")
+
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])
+ # 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 oth_node, tgt_node:
+
+ for node in nodes:
res = results[node]
- msg = res.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Error checking node %s: %s" % (node, msg))
- if my_vg not in res.payload:
- 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:
+ 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 node in tgt_node, oth_node:
- info("checking disk/%d on %s" % (idx, node))
- cfg.SetDiskID(dev, node)
+
+ for node in nodes:
+ self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
+ self.cfg.SetDiskID(dev, node)
+
result = self.rpc.call_blockdev_find(node, dev)
- msg = result.RemoteFailMsg()
- if not msg and not result.payload:
- msg = "disk not found"
- if msg:
+
+ 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))
- # 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:
+ def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
+ for idx, dev in enumerate(self.instance.disks):
+ if idx not in self.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))
- # 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:
+ 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
- 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,
+
+ 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)
- 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)
+ _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:
+
+ 1. for each disk to be replaced:
+
+ 1. create new LVs on the target node with unique names
+ 1. detach old LVs from the drbd device
+ 1. rename old LVs to name_replaced.<time_t>
+ 1. rename new LVs to old LVs
+ 1. attach the new LVs (with the old names now) to the drbd device
+
+ 1. wait for sync across all devices
+
+ 1. for each modified disk:
+
+ 1. remove old LVs (which have the name name_replaces.<time_t>)
+
+ Failures are not very well handled.
+
+ """
+ steps_total = 6
+
+ # Step: check device activation
+ 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.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.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)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't detach drbd from local storage on node"
- " %s for device %s: %s" %
- (tgt_node, dev.iv_name, msg))
+ 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:
- result = self.rpc.call_blockdev_find(tgt_node, to_ren)
- if not result.RemoteFailMsg() and result.payload:
+ result = self.rpc.call_blockdev_find(self.target_node, to_ren)
+ if not result.fail_msg and result.payload:
# device exists
- rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
+ rename_old_to_new.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)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't rename old LVs on node %s: %s" %
- (tgt_node, msg))
- # 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)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't rename new LVs on node %s: %s" %
- (tgt_node, msg))
+ 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)
- msg = result.RemoteFailMsg()
+ # 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:
- msg = self.rpc.call_blockdev_remove(tgt_node, new_lv).RemoteFailMsg()
- if msg:
- warning("Can't rollback device %s: %s", dev, msg,
- hint="cleanup manually the unused logical volumes")
+ 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)
- msg = result.RemoteFailMsg()
- if not msg and not result.payload:
- msg = "disk not found"
- if msg:
- 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)
+ # 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)
- msg = self.rpc.call_blockdev_remove(tgt_node, lv).RemoteFailMsg()
- if msg:
- warning("Can't remove old LV: %s" % msg,
- 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]
- msg = res.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Error checking node %s: %s" % (node, msg))
- if my_vg not in res.payload:
- 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)
- msg = result.RemoteFailMsg()
- if not msg and not result.payload:
- msg = "disk not found"
- if msg:
- raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
- (idx, pri_node, msg))
+ 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)
+ _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
+ _GetInstanceInfoText(self.instance), False)
except errors.GenericError:
- self.cfg.ReleaseDRBDMinors(instance.name)
+ 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)
- msg = self.rpc.call_blockdev_shutdown(old_node, dev).RemoteFailMsg()
+ # 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:
- warning("Failed to shutdown drbd for disk/%d on old node: %s" %
- (idx, msg),
- hint="Please cleanup this device manually as soon as possible")
+ 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)
+ 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)
- msg = result.RemoteFailMsg()
- if not msg and not result.payload:
- msg = "disk not found"
- if msg:
- raise errors.OpExecError("Can't find DRBD device disk/%d: %s" %
- (idx, msg))
- if result.payload[5]:
- raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
+ 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"))
- 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)
- msg = self.rpc.call_blockdev_remove(old_node, lv).RemoteFailMsg()
- if msg:
- warning("Can't remove LV on old secondary: %s", msg,
- hint="Cleanup stale volumes by hand")
+ # 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)
- def Exec(self, feedback_fn):
- """Execute disk replacement.
-
- This dispatches the disk replacement to the appropriate handler.
-
- """
- 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 not instance.admin_up:
- _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 not instance.admin_up:
- _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]
- msg = info.RemoteFailMsg()
- if msg:
- raise errors.OpPrereqError("Cannot get current information"
- " from node %s:" % (node, msg))
+ 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"
for node in instance.all_nodes:
self.cfg.SetDiskID(disk, node)
result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Grow request failed to node %s: %s" %
- (node, msg))
+ 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 dev_pstatus.offline:
dev_pstatus = None
else:
- msg = dev_pstatus.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't compute disk status for %s: %s" %
- (instance.name, msg))
+ dev_pstatus.Raise("Can't compute disk status for %s" % instance.name)
dev_pstatus = dev_pstatus.payload
else:
dev_pstatus = None
if dev_sstatus.offline:
dev_sstatus = None
else:
- msg = dev_sstatus.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't compute disk status for %s: %s" %
- (instance.name, msg))
+ 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)
- msg = remote_info.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Error checking node %s: %s" %
- (instance.primary_node, msg))
+ 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"
"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,
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_ACCESS_SET:
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)
nic_bridge = nic_dict.get('bridge', None)
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")
+ 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:
"""Return the new params dict for the given params.
@type old_params: dict
- @type old_params: old parameters
+ @param old_params: old parameters
@type update_dict: dict
- @type update_dict: dict containing new parameter values,
- or constants.VALUE_DEFAULT to reset the
- parameter to its default value
+ @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
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
nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
instance.hypervisor)
pninfo = nodeinfo[pnode]
- msg = pninfo.RemoteFailMsg()
+ 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: %s" %
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.RemoteFailMsg():
+ elif instance_info.fail_msg:
self.warn.append("Can't get instance runtime information: %s" %
- instance_info.RemoteFailMsg())
+ instance_info.fail_msg)
else:
if instance_info.payload:
current_mem = int(instance_info.payload['memory'])
for node, nres in nodeinfo.items():
if node not in instance.secondary_nodes:
continue
- msg = nres.RemoteFailMsg()
+ msg = nres.fail_msg
if msg:
self.warn.append("Can't get info from secondary node %s: %s" %
(node, msg))
if new_nic_mode == constants.NIC_MODE_BRIDGED:
nic_bridge = new_filled_nic_params[constants.NIC_LINK]
- result = self.rpc.call_bridges_exist(pnode, [nic_bridge])
- msg = result.RemoteFailMsg()
+ 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:
" an instance")
ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
ins_l = ins_l[pnode]
- msg = ins_l.RemoteFailMsg()
+ msg = ins_l.fail_msg
if msg:
raise errors.OpPrereqError("Can't contact node %s: %s" %
(pnode, msg))
device_idx = len(instance.disks)
for node, disk in device.ComputeNodeTree(instance.primary_node):
self.cfg.SetDiskID(disk, node)
- msg = self.rpc.call_blockdev_remove(node, disk).RemoteFailMsg()
+ 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)
rpcresult = self.rpc.call_export_list(self.nodes)
result = {}
for node in rpcresult:
- if rpcresult[node].RemoteFailMsg():
+ if rpcresult[node].fail_msg:
result[node] = False
else:
result[node] = rpcresult[node].payload
# 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):
if self.op.shutdown:
# shutdown the instance, but not the disks
result = self.rpc.call_instance_shutdown(src_node, instance)
- msg = result.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Could not shutdown instance %s on"
- " node %s: %s" %
- (instance.name, src_node, msg))
+ 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:
+ 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.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
- self.LogWarning("Could not snapshot block device %s on node %s: %s",
- disk.logical_id[1], src_node, 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)
finally:
if self.op.shutdown and instance.admin_up:
result = self.rpc.call_instance_start(src_node, instance, None, None)
- msg = result.RemoteFailMsg()
+ 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)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
- self.LogWarning("Could not export block device %s from node %s to"
- " node %s: %s", dev.logical_id[1], src_node,
- dst_node.name, msg)
- msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
+ 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 block device %s from node"
- " %s: %s", dev.logical_id[1], src_node, 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)
- msg = result.RemoteFailMsg()
+ 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)
if nodelist:
exportlist = self.rpc.call_export_list(nodelist)
for node in exportlist:
- if exportlist[node].RemoteFailMsg():
+ if exportlist[node].fail_msg:
continue
if iname in exportlist[node].payload:
- msg = self.rpc.call_export_remove(node, iname).RemoteFailMsg()
+ 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: %s", iname, node, msg)
exportlist = self.rpc.call_export_list(locked_nodes)
found = False
for node in exportlist:
- msg = exportlist[node].RemoteFailMsg()
+ 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].payload:
found = True
result = self.rpc.call_export_remove(node, instance_name)
- msg = result.RemoteFailMsg()
+ msg = result.fail_msg
if msg:
logging.error("Could not remove export for instance %s"
" on node %s: %s", instance_name, node, msg)
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 = {
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)
+ 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)
}
if not ninfo.offline:
- msg = nresult.RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't get data for node %s: %s" %
- (nname, msg))
- msg = node_iinfo[nname].RemoteFailMsg()
- if msg:
- raise errors.OpExecError("Can't get node instance info"
- " from node %s: %s" % (nname, msg))
+ 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']:
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()
"""
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),
projects / ganeti-local / blobdiff