import os
import os.path
-import sha
import time
import tempfile
import re
if nics:
nic_count = len(nics)
- for idx, (ip, bridge, mac) in enumerate(nics):
+ for idx, (ip, mac, mode, link) in enumerate(nics):
if ip is None:
ip = ""
env["INSTANCE_NIC%d_IP" % idx] = ip
- env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
env["INSTANCE_NIC%d_MAC" % idx] = mac
+ env["INSTANCE_NIC%d_MODE" % idx] = mode
+ env["INSTANCE_NIC%d_LINK" % idx] = link
+ if mode == constants.NIC_MODE_BRIDGED:
+ env["INSTANCE_NIC%d_BRIDGE" % idx] = link
else:
nic_count = 0
return env
+def _PreBuildNICHooksList(lu, nics):
+ """Build a list of nic information tuples.
+
+ This list is suitable to be passed to _BuildInstanceHookEnv.
+
+ @type lu: L{LogicalUnit}
+ @param lu: the logical unit on whose behalf we execute
+ @type nics: list of L{objects.NIC}
+ @param nics: list of nics to convert to hooks tuples
+
+ """
+ hooks_nics = []
+ c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
+ for nic in nics:
+ ip = nic.ip
+ mac = nic.mac
+ filled_params = objects.FillDict(c_nicparams, nic.nicparams)
+ mode = filled_params[constants.NIC_MODE]
+ link = filled_params[constants.NIC_LINK]
+ hooks_nics.append((ip, mac, mode, link))
+ return hooks_nics
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
"""Builds instance related env variables for hooks from an object.
'status': instance.admin_up,
'memory': bep[constants.BE_MEMORY],
'vcpus': bep[constants.BE_VCPUS],
- 'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
+ 'nics': _PreBuildNICHooksList(lu, instance.nics),
'disk_template': instance.disk_template,
'disks': [(disk.size, disk.mode) for disk in instance.disks],
}
(mc_now, mc_max))
-def _CheckInstanceBridgesExist(lu, instance):
+def _CheckNicsBridgesExist(lu, target_nics, target_node,
+ profile=constants.PP_DEFAULT):
+ """Check that the brigdes needed by a list of nics exist.
+
+ """
+ c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
+ paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
+ for nic in target_nics]
+ brlist = [params[constants.NIC_LINK] for params in paramslist
+ if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
+ if brlist:
+ result = lu.rpc.call_bridges_exist(target_node, brlist)
+ msg = result.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Error checking bridges on destination node"
+ " '%s': %s" % (target_node, msg))
+
+
+def _CheckInstanceBridgesExist(lu, instance, node=None):
"""Check that the brigdes needed by an instance exist.
"""
- # check bridges existance
- brlist = [nic.bridge for nic in instance.nics]
- result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("One or more target bridges %s does not"
- " exist on destination node '%s'" %
- (brlist, instance.primary_node))
+ if node is None:
+ node=instance.primary_node
+ _CheckNicsBridgesExist(lu, instance.nics, node)
class LUDestroyCluster(NoHooksLU):
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].RemoteFailMsg()
+ 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():
- instance = instanceinfo[instance]
- node_drbd[minor] = (instance.name, instance.admin_up)
+ if instance not in instanceinfo:
+ feedback_fn(" - ERROR: ghost instance '%s' in temporary DRBD map" %
+ instance)
+ # ghost instance should not be running, but otherwise we
+ # don't give double warnings (both ghost instance and
+ # unallocated minor in use)
+ node_drbd[minor] = (instance, False)
+ else:
+ instance = instanceinfo[instance]
+ node_drbd[minor] = (instance.name, instance.admin_up)
result = self._VerifyNode(node_i, file_names, local_checksums,
nresult, feedback_fn, master_files,
node_drbd, vg_name)
}
# FIXME: devise a free space model for file based instances as well
if vg_name is not None:
+ if (constants.NV_VGLIST not in nresult or
+ vg_name not in nresult[constants.NV_VGLIST]):
+ feedback_fn(" - ERROR: node %s didn't return data for the"
+ " volume group '%s' - it is either missing or broken" %
+ (node, vg_name))
+ bad = True
+ continue
node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
- except ValueError:
- feedback_fn(" - ERROR: invalid value returned from node %s" % (node,))
+ except (ValueError, KeyError):
+ feedback_fn(" - ERROR: invalid nodeinfo value returned"
+ " from node %s" % (node,))
bad = True
continue
def Exec(self, feedback_fn):
"""Verify integrity of cluster disks.
+ @rtype: tuple of three items
+ @return: a tuple of (dict of node-to-node_error, list of instances
+ which need activate-disks, dict of instance: (node, volume) for
+ missing volumes
+
"""
- result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
+ result = res_nodes, res_instances, res_missing = {}, [], {}
vg_name = self.cfg.GetVGName()
nodes = utils.NiceSort(self.cfg.GetNodeList())
to_act = set()
for node in nodes:
# node_volume
- lvs = node_lvs[node]
- if lvs.failed:
- if not lvs.offline:
- self.LogWarning("Connection to node %s failed: %s" %
- (node, lvs.data))
+ node_res = node_lvs[node]
+ if node_res.offline:
continue
- lvs = lvs.data
- if isinstance(lvs, basestring):
- logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
- res_nlvm[node] = lvs
- elif not isinstance(lvs, dict):
- logging.warning("Connection to node %s failed or invalid data"
- " returned", node)
- res_nodes.append(node)
+ msg = node_res.RemoteFailMsg()
+ if msg:
+ logging.warning("Error enumerating LVs on node %s: %s", node, msg)
+ res_nodes[node] = msg
continue
- for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
+ lvs = node_res.payload
+ for lv_name, (_, lv_inactive, lv_online) in lvs.items():
inst = nv_dict.pop((node, lv_name), None)
if (not lv_online and inst is not None
and inst.name not in res_instances):
result = self.rpc.call_upload_file(node_list,
constants.SSH_KNOWN_HOSTS_FILE)
for to_node, to_result in result.iteritems():
- if to_result.failed or not to_result.data:
- logging.error("Copy of file %s to node %s failed",
- constants.SSH_KNOWN_HOSTS_FILE, to_node)
+ msg = to_result.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)
finally:
result = self.rpc.call_node_start_master(master, False)
- if result.failed or not result.data:
+ msg = result.RemoteFailMsg()
+ 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):
_OP_REQP = []
REQ_BGL = False
- def CheckParameters(self):
+ def CheckArguments(self):
"""Check parameters
"""
if self.op.candidate_pool_size is not None:
try:
self.op.candidate_pool_size = int(self.op.candidate_pool_size)
- except ValueError, err:
+ except (ValueError, TypeError), err:
raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
str(err))
if self.op.candidate_pool_size < 1:
if self.op.vg_name:
vglist = self.rpc.call_vg_list(node_list)
for node in node_list:
- if vglist[node].failed:
+ msg = vglist[node].RemoteFailMsg()
+ if msg:
# ignoring down node
- self.LogWarning("Node %s unreachable/error, ignoring" % node)
+ self.LogWarning("Error while gathering data on node %s"
+ " (ignoring node): %s", node, msg)
continue
- vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
+ vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
self.op.vg_name,
constants.MIN_VG_SIZE)
if vgstatus:
(node, vgstatus))
self.cluster = cluster = self.cfg.GetClusterInfo()
- # validate beparams changes
+ # validate params changes
if self.op.beparams:
utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
- self.new_beparams = cluster.FillDict(
- cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
+ self.new_beparams = objects.FillDict(
+ cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
+
+ if self.op.nicparams:
+ utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
+ self.new_nicparams = objects.FillDict(
+ cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
+ objects.NIC.CheckParameterSyntax(self.new_nicparams)
# hypervisor list/parameters
- self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
+ self.new_hvparams = objects.FillDict(cluster.hvparams, {})
if self.op.hvparams:
if not isinstance(self.op.hvparams, dict):
raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
"""
if self.op.vg_name is not None:
- if self.op.vg_name != self.cfg.GetVGName():
- self.cfg.SetVGName(self.op.vg_name)
+ new_volume = self.op.vg_name
+ if not new_volume:
+ new_volume = None
+ if new_volume != self.cfg.GetVGName():
+ self.cfg.SetVGName(new_volume)
else:
feedback_fn("Cluster LVM configuration already in desired"
" state, not changing")
if self.op.enabled_hypervisors is not None:
self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
if self.op.beparams:
- self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
+ self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
+ if self.op.nicparams:
+ self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
+
if self.op.candidate_pool_size is not None:
self.cluster.candidate_pool_size = self.op.candidate_pool_size
_AdjustCandidatePool(self)
+def _RedistributeAncillaryFiles(lu, additional_nodes=None):
+ """Distribute additional files which are part of the cluster configuration.
+
+ ConfigWriter takes care of distributing the config and ssconf files, but
+ there are more files which should be distributed to all nodes. This function
+ makes sure those are copied.
+
+ @param lu: calling logical unit
+ @param additional_nodes: list of nodes not in the config to distribute to
+
+ """
+ # 1. Gather target nodes
+ myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
+ dist_nodes = lu.cfg.GetNodeList()
+ if additional_nodes is not None:
+ dist_nodes.extend(additional_nodes)
+ if myself.name in dist_nodes:
+ dist_nodes.remove(myself.name)
+ # 2. Gather files to distribute
+ dist_files = set([constants.ETC_HOSTS,
+ constants.SSH_KNOWN_HOSTS_FILE,
+ constants.RAPI_CERT_FILE,
+ constants.RAPI_USERS_FILE,
+ ])
+
+ enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
+ for hv_name in enabled_hypervisors:
+ hv_class = hypervisor.GetHypervisor(hv_name)
+ dist_files.update(hv_class.GetAncillaryFiles())
+
+ # 3. Perform the files upload
+ for fname in dist_files:
+ if os.path.exists(fname):
+ result = lu.rpc.call_upload_file(dist_nodes, fname)
+ for to_node, to_result in result.items():
+ msg = to_result.RemoteFailMsg()
+ if msg:
+ msg = ("Copy of file %s to node %s failed: %s" %
+ (fname, to_node, msg))
+ lu.proc.LogWarning(msg)
+
+
class LURedistributeConfig(NoHooksLU):
"""Force the redistribution of cluster configuration.
"""
self.cfg.Update(self.cfg.GetClusterInfo())
+ _RedistributeAncillaryFiles(self)
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
done = True
cumul_degraded = False
rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
- if rstats.failed or not rstats.data:
- lu.LogWarning("Can't get any data from node %s", node)
+ msg = rstats.RemoteFailMsg()
+ if msg:
+ lu.LogWarning("Can't get any data from node %s: %s", node, msg)
retries += 1
if retries >= 10:
raise errors.RemoteError("Can't contact node %s for mirror data,"
" aborting." % node)
time.sleep(6)
continue
- rstats = rstats.data
+ rstats = rstats.payload
retries = 0
for i, mstat in enumerate(rstats):
if mstat is None:
selected=self.op.output_fields)
# Lock all nodes, in shared mode
+ # Temporary removal of locks, should be reverted later
+ # TODO: reintroduce locks when they are lighter-weight
self.needed_locks = {}
- self.share_locks[locking.LEVEL_NODE] = 1
- self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
+ #self.share_locks[locking.LEVEL_NODE] = 1
+ #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
def CheckPrereq(self):
"""Check prerequisites.
"""
all_os = {}
+ # we build here the list of nodes that didn't fail the RPC (at RPC
+ # level), so that nodes with a non-responding node daemon don't
+ # make all OSes invalid
+ good_nodes = [node_name for node_name in rlist
+ if not rlist[node_name].failed]
for node_name, nr in rlist.iteritems():
if nr.failed or not nr.data:
continue
# build a list of nodes for this os containing empty lists
# for each node in node_list
all_os[os_obj.name] = {}
- for nname in node_list:
+ for nname in good_nodes:
all_os[os_obj.name][nname] = []
all_os[os_obj.name][node_name].append(os_obj)
return all_os
"""Compute the list of OSes.
"""
- node_list = self.acquired_locks[locking.LEVEL_NODE]
- valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
- if node in node_list]
+ valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
node_data = self.rpc.call_os_diagnose(valid_nodes)
if node_data == False:
raise errors.OpExecError("Can't gather the list of OSes")
self.cfg.GetHypervisorType())
for name in nodenames:
nodeinfo = node_data[name]
- if not nodeinfo.failed and nodeinfo.data:
- nodeinfo = nodeinfo.data
+ if not nodeinfo.RemoteFailMsg() and nodeinfo.payload:
+ nodeinfo = nodeinfo.payload
fn = utils.TryConvert
live_data[name] = {
"mtotal": fn(int, nodeinfo.get('memory_total', None)),
" new node: %s" % msg)
# Add node to our /etc/hosts, and add key to known_hosts
- utils.AddHostToEtcHosts(new_node.name)
+ if self.cfg.GetClusterInfo().modify_etc_hosts:
+ utils.AddHostToEtcHosts(new_node.name)
if new_node.secondary_ip != new_node.primary_ip:
result = self.rpc.call_node_has_ip_address(new_node.name,
new_node.secondary_ip)
- if result.failed or not result.data:
+ msg = result.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Failure checking secondary ip"
+ " on node %s: %s" % (new_node.name, msg))
+ if not result.payload:
raise errors.OpExecError("Node claims it doesn't have the secondary ip"
" you gave (%s). Please fix and re-run this"
" command." % new_node.secondary_ip)
result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
self.cfg.GetClusterName())
for verifier in node_verify_list:
- if result[verifier].failed or not result[verifier].data:
- raise errors.OpExecError("Cannot communicate with %s's node daemon"
- " for remote verification" % verifier)
- if result[verifier].data['nodelist']:
- for failed in result[verifier].data['nodelist']:
+ msg = result[verifier].RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Cannot communicate with node %s: %s" %
+ (verifier, msg))
+ 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.")
- # Distribute updated /etc/hosts and known_hosts to all nodes,
- # including the node just added
- myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
- dist_nodes = self.cfg.GetNodeList()
- if not self.op.readd:
- dist_nodes.append(node)
- if myself.name in dist_nodes:
- dist_nodes.remove(myself.name)
-
- logging.debug("Copying hosts and known_hosts to all nodes")
- for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
- result = self.rpc.call_upload_file(dist_nodes, fname)
- for to_node, to_result in result.iteritems():
- if to_result.failed or not to_result.data:
- logging.error("Copy of file %s to node %s failed", fname, to_node)
-
- to_copy = []
- enabled_hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
- if constants.HTS_COPY_VNC_PASSWORD.intersection(enabled_hypervisors):
- to_copy.append(constants.VNC_PASSWORD_FILE)
-
- for fname in to_copy:
- result = self.rpc.call_upload_file([node], fname)
- if result[node].failed or not result[node]:
- logging.error("Could not copy file %s to node %s", fname, node)
-
if self.op.readd:
+ _RedistributeAncillaryFiles(self)
self.context.ReaddNode(new_node)
else:
+ _RedistributeAncillaryFiles(self, additional_nodes=[node])
self.context.AddNode(new_node)
((node.offline and not self.op.offline == False) or
(node.drained and not self.op.drained == False))):
raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
- " to master_candidate")
+ " to master_candidate" % node.name)
return
return result
+class LUPowercycleNode(NoHooksLU):
+ """Powercycles a node.
+
+ """
+ _OP_REQP = ["node_name", "force"]
+ REQ_BGL = False
+
+ def CheckArguments(self):
+ node_name = self.cfg.ExpandNodeName(self.op.node_name)
+ if node_name is None:
+ raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
+ self.op.node_name = node_name
+ if node_name == self.cfg.GetMasterNode() and not self.op.force:
+ raise errors.OpPrereqError("The node is the master and the force"
+ " parameter was not set")
+
+ def ExpandNames(self):
+ """Locking for PowercycleNode.
+
+ This is a last-resource option and shouldn't block on other
+ jobs. Therefore, we grab no locks.
+
+ """
+ self.needed_locks = {}
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This LU has no prereqs.
+
+ """
+ pass
+
+ def Exec(self, feedback_fn):
+ """Reboots a node.
+
+ """
+ result = self.rpc.call_node_powercycle(self.op.node_name,
+ self.cfg.GetHypervisorType())
+ msg = result.RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Failed to schedule the reboot: %s" % msg)
+ return result.payload
+
+
class LUQueryClusterInfo(NoHooksLU):
"""Query cluster configuration.
"hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
for hypervisor in cluster.enabled_hypervisors]),
"beparams": cluster.beparams,
+ "nicparams": cluster.nicparams,
"candidate_pool_size": cluster.candidate_pool_size,
+ "master_netdev": cluster.master_netdev,
+ "volume_group_name": cluster.volume_group_name,
+ "file_storage_dir": cluster.file_storage_dir,
}
return result
_ShutdownInstanceDisks.
"""
- ins_l = lu.rpc.call_instance_list([instance.primary_node],
- [instance.hypervisor])
- ins_l = ins_l[instance.primary_node]
- if ins_l.failed or not isinstance(ins_l.data, list):
- raise errors.OpExecError("Can't contact node '%s'" %
- instance.primary_node)
-
- if instance.name in ins_l.data:
+ pnode = instance.primary_node
+ ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])
+ ins_l = ins_l[pnode]
+ msg = ins_l.RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Can't contact node %s: %s" % (pnode, msg))
+
+ if instance.name in ins_l.payload:
raise errors.OpExecError("Instance is running, can't shutdown"
" block devices.")
"""
nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
- nodeinfo[node].Raise()
- free_mem = nodeinfo[node].data.get('memory_free')
+ msg = nodeinfo[node].RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Can't get data from node %s: %s" % (node, msg))
+ free_mem = nodeinfo[node].payload.get('memory_free', None)
if not isinstance(free_mem, int):
raise errors.OpPrereqError("Can't compute free memory on node %s, result"
- " was '%s'" % (node, free_mem))
+ " was '%s'" % (node, free_mem))
if requested > free_mem:
raise errors.OpPrereqError("Not enough memory on node %s for %s:"
- " needed %s MiB, available %s MiB" %
- (node, reason, requested, free_mem))
+ " needed %s MiB, available %s MiB" %
+ (node, reason, requested, free_mem))
class LUStartupInstance(LogicalUnit):
assert self.instance is not None, \
"Cannot retrieve locked instance %s" % self.op.instance_name
+ # extra beparams
+ self.beparams = getattr(self.op, "beparams", {})
+ if self.beparams:
+ if not isinstance(self.beparams, dict):
+ raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
+ " dict" % (type(self.beparams), ))
+ # fill the beparams dict
+ utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
+ self.op.beparams = self.beparams
+
+ # extra hvparams
+ self.hvparams = getattr(self.op, "hvparams", {})
+ if self.hvparams:
+ if not isinstance(self.hvparams, dict):
+ raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
+ " dict" % (type(self.hvparams), ))
+
+ # check hypervisor parameter syntax (locally)
+ cluster = self.cfg.GetClusterInfo()
+ utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
+ filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
+ instance.hvparams)
+ filled_hvp.update(self.hvparams)
+ hv_type = hypervisor.GetHypervisor(instance.hypervisor)
+ hv_type.CheckParameterSyntax(filled_hvp)
+ _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
+ self.op.hvparams = self.hvparams
+
_CheckNodeOnline(self, instance.primary_node)
bep = self.cfg.GetClusterInfo().FillBE(instance)
# check bridges existance
_CheckInstanceBridgesExist(self, instance)
- _CheckNodeFreeMemory(self, instance.primary_node,
- "starting instance %s" % instance.name,
- bep[constants.BE_MEMORY], instance.hypervisor)
+ remote_info = self.rpc.call_instance_info(instance.primary_node,
+ instance.name,
+ instance.hypervisor)
+ msg = remote_info.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Error checking node %s: %s" %
+ (instance.primary_node, msg))
+ if not remote_info.payload: # not running already
+ _CheckNodeFreeMemory(self, instance.primary_node,
+ "starting instance %s" % instance.name,
+ bep[constants.BE_MEMORY], instance.hypervisor)
def Exec(self, feedback_fn):
"""Start the instance.
_StartInstanceDisks(self, instance, force)
- result = self.rpc.call_instance_start(node_current, instance)
+ result = self.rpc.call_instance_start(node_current, instance,
+ self.hvparams, self.beparams)
msg = result.RemoteFailMsg()
if msg:
_ShutdownInstanceDisks(self, instance)
" full reboot: %s" % msg)
_ShutdownInstanceDisks(self, instance)
_StartInstanceDisks(self, instance, ignore_secondaries)
- result = self.rpc.call_instance_start(node_current, instance)
+ result = self.rpc.call_instance_start(node_current, instance, None, None)
msg = result.RemoteFailMsg()
if msg:
_ShutdownInstanceDisks(self, instance)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- if remote_info.failed or remote_info.data:
+ msg = remote_info.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Error checking node %s: %s" %
+ (instance.primary_node, msg))
+ if remote_info.payload:
raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
(self.op.instance_name,
instance.primary_node))
_StartInstanceDisks(self, inst, None)
try:
feedback_fn("Running the instance OS create scripts...")
- result = self.rpc.call_instance_os_add(inst.primary_node, inst)
+ 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"
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- remote_info.Raise()
- if remote_info.data:
+ msg = remote_info.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Error checking node %s: %s" %
+ (instance.primary_node, msg))
+ if remote_info.payload:
raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
(self.op.instance_name,
instance.primary_node))
if result.offline:
# offline nodes will be in both lists
off_nodes.append(name)
- if result.failed:
+ if result.failed or result.RemoteFailMsg():
bad_nodes.append(name)
else:
- if result.data:
- live_data.update(result.data)
- # else no instance is alive
+ if result.payload:
+ live_data.update(result.payload)
+ # else no instance is alive
else:
live_data = dict([(name, {}) for name in instance_names])
_CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
instance.name, bep[constants.BE_MEMORY],
instance.hypervisor)
-
# check bridge existance
- brlist = [nic.bridge for nic in instance.nics]
- result = self.rpc.call_bridges_exist(target_node, brlist)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("One or more target bridges %s does not"
- " exist on destination node '%s'" %
- (brlist, target_node))
+ _CheckInstanceBridgesExist(self, instance, node=target_node)
def Exec(self, feedback_fn):
"""Failover an instance.
raise errors.OpExecError("Can't activate the instance's disks")
feedback_fn("* starting the instance on the target node")
- result = self.rpc.call_instance_start(target_node, instance)
+ result = self.rpc.call_instance_start(target_node, instance, None, None)
msg = result.RemoteFailMsg()
if msg:
_ShutdownInstanceDisks(self, instance)
instance.hypervisor)
# check bridge existance
- brlist = [nic.bridge for nic in instance.nics]
- result = self.rpc.call_bridges_exist(target_node, brlist)
- if result.failed or not result.data:
- raise errors.OpPrereqError("One or more target bridges %s does not"
- " exist on destination node '%s'" %
- (brlist, target_node))
+ _CheckInstanceBridgesExist(self, instance, node=target_node)
if not self.op.cleanup:
_CheckNodeNotDrained(self, target_node)
" a bad state)")
ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
for node, result in ins_l.items():
- result.Raise()
- if not isinstance(result.data, list):
- raise errors.OpExecError("Can't contact node '%s'" % node)
+ msg = result.RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Can't contact node %s: %s" % (node, msg))
- runningon_source = instance.name in ins_l[source_node].data
- runningon_target = instance.name in ins_l[target_node].data
+ runningon_source = instance.name in ins_l[source_node].payload
+ runningon_target = instance.name in ins_l[target_node].payload
if runningon_source and runningon_target:
raise errors.OpExecError("Instance seems to be running on two nodes,"
continue
msg = info.RemoteFailMsg()
if msg:
- raise errors.OpPrereqError("Hypervisor parameter validation failed:"
- " %s" % msg)
+ raise errors.OpPrereqError("Hypervisor parameter validation"
+ " failed on node %s: %s" % (node, msg))
class LUCreateInstance(LogicalUnit):
# check hypervisor parameter syntax (locally)
utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
- filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
+ filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
self.op.hvparams)
hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
hv_type.CheckParameterSyntax(filled_hvp)
# fill and remember the beparams dict
utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
- self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
+ self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
self.op.beparams)
#### instance parameters check
# NIC buildup
self.nics = []
- for nic in self.op.nics:
+ for idx, nic in enumerate(self.op.nics):
+ nic_mode_req = nic.get("mode", None)
+ nic_mode = nic_mode_req
+ if nic_mode is None:
+ nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
+
+ # in routed mode, for the first nic, the default ip is 'auto'
+ if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
+ default_ip_mode = constants.VALUE_AUTO
+ else:
+ default_ip_mode = constants.VALUE_NONE
+
# ip validity checks
- ip = nic.get("ip", None)
- if ip is None or ip.lower() == "none":
+ ip = nic.get("ip", default_ip_mode)
+ if ip is None or ip.lower() == constants.VALUE_NONE:
nic_ip = None
elif ip.lower() == constants.VALUE_AUTO:
nic_ip = hostname1.ip
" like a valid IP" % ip)
nic_ip = ip
+ # TODO: check the ip for uniqueness !!
+ if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
+ raise errors.OpPrereqError("Routed nic mode requires an ip address")
+
# MAC address verification
mac = nic.get("mac", constants.VALUE_AUTO)
if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
mac)
# bridge verification
bridge = nic.get("bridge", None)
- if bridge is None:
- bridge = self.cfg.GetDefBridge()
- self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
+ link = nic.get("link", None)
+ if bridge and link:
+ raise errors.OpPrereqError("Cannot pass 'bridge' and 'link' at the same time")
+ elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
+ raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic")
+ elif bridge:
+ link = bridge
+
+ nicparams = {}
+ if nic_mode_req:
+ nicparams[constants.NIC_MODE] = nic_mode_req
+ if link:
+ nicparams[constants.NIC_LINK] = link
+
+ check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
+ nicparams)
+ objects.NIC.CheckParameterSyntax(check_params)
+ self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
# disk checks/pre-build
self.disks = []
os_type=self.op.os_type,
memory=self.be_full[constants.BE_MEMORY],
vcpus=self.be_full[constants.BE_VCPUS],
- nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
+ nics=_PreBuildNICHooksList(self, self.nics),
disk_template=self.op.disk_template,
disks=[(d["size"], d["mode"]) for d in self.disks],
))
src_path = self.op.src_path
if src_node is None:
- exp_list = self.rpc.call_export_list(
- self.acquired_locks[locking.LEVEL_NODE])
+ locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
+ exp_list = self.rpc.call_export_list(locked_nodes)
found = False
for node in exp_list:
- if not exp_list[node].failed and src_path in exp_list[node].data:
+ if exp_list[node].RemoteFailMsg():
+ continue
+ if src_path in exp_list[node].payload:
found = True
self.op.src_node = src_node = node
self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
_CheckNodeOnline(self, src_node)
result = self.rpc.call_export_info(src_node, src_path)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("No export found in dir %s" % src_path)
+ msg = result.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("No export or invalid export found in"
+ " dir %s: %s" % (src_path, msg))
- export_info = result.data
+ export_info = objects.SerializableConfigParser.Loads(str(result.payload))
if not export_info.has_section(constants.INISECT_EXP):
raise errors.ProgrammerError("Corrupted export config")
self.op.hypervisor)
for node in nodenames:
info = nodeinfo[node]
- info.Raise()
- info = info.data
- if not info:
+ msg = info.RemoteFailMsg()
+ if msg:
raise errors.OpPrereqError("Cannot get current information"
- " from node '%s'" % node)
+ " from node %s: %s" % (node, msg))
+ info = info.payload
vg_free = info.get('vg_free', None)
if not isinstance(vg_free, int):
raise errors.OpPrereqError("Can't compute free disk space on"
" node %s" % node)
- if req_size > info['vg_free']:
+ if req_size > vg_free:
raise errors.OpPrereqError("Not enough disk space on target node %s."
" %d MB available, %d MB required" %
- (node, info['vg_free'], req_size))
+ (node, vg_free, req_size))
_CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
raise errors.OpPrereqError("OS '%s' not in supported os list for"
" primary node" % self.op.os_type)
- # bridge check on primary node
- bridges = [n.bridge for n in self.nics]
- result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
- result.Raise()
- if not result.data:
- raise errors.OpPrereqError("One of the target bridges '%s' does not"
- " exist on destination node '%s'" %
- (",".join(bridges), pnode.name))
+ _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
# memory check on primary node
if self.op.start:
if iobj.disk_template != constants.DT_DISKLESS:
if self.op.mode == constants.INSTANCE_CREATE:
feedback_fn("* running the instance OS create scripts...")
- result = self.rpc.call_instance_os_add(pnode_name, iobj)
+ 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"
import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
src_node, src_images,
cluster_name)
- import_result.Raise()
- for idx, result in enumerate(import_result.data):
- if not result:
- self.LogWarning("Could not import the image %s for instance"
- " %s, disk %d, on node %s" %
- (src_images[idx], instance, idx, pnode_name))
+ msg = import_result.RemoteFailMsg()
+ if msg:
+ self.LogWarning("Error while importing the disk images for instance"
+ " %s on node %s: %s" % (instance, pnode_name, msg))
else:
# also checked in the prereq part
raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
self.cfg.Update(iobj)
logging.info("Starting instance %s on node %s", instance, pnode_name)
feedback_fn("* starting instance...")
- result = self.rpc.call_instance_start(pnode_name, iobj)
+ 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)
node_insts = self.rpc.call_instance_list([node],
[instance.hypervisor])[node]
- node_insts.Raise()
+ msg = node_insts.RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Can't get node information from %s: %s" %
+ (node, msg))
- if instance.name not in node_insts.data:
+ if instance.name not in node_insts.payload:
raise errors.OpExecError("Instance %s is not running." % instance.name)
logging.debug("Connecting to console of %s on %s", instance.name, node)
raise errors.OpExecError("Can't list volume groups on the nodes")
for node in oth_node, tgt_node:
res = results[node]
- if res.failed or not res.data or my_vg not in res.data:
+ 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):
for dev, old_lvs, new_lvs in iv_names.itervalues():
info("detaching %s drbd from local storage" % dev.iv_name)
result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
- result.Raise()
- if not result.data:
+ msg = result.RemoteFailMsg()
+ if msg:
raise errors.OpExecError("Can't detach drbd from local storage on node"
- " %s for device %s" % (tgt_node, dev.iv_name))
+ " %s for device %s: %s" %
+ (tgt_node, dev.iv_name, msg))
#dev.children = []
#cfg.Update(instance)
info("renaming the old LVs on the target node")
result = self.rpc.call_blockdev_rename(tgt_node, rlist)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
+ 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)
- result.Raise()
- if not result.data:
- raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
+ msg = result.RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Can't rename new LVs on node %s: %s" %
+ (tgt_node, msg))
for old, new in zip(old_lvs, new_lvs):
new.logical_id = old.logical_id
# now that the new lvs have the old name, we can add them to the device
info("adding new mirror component on %s" % tgt_node)
result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
- if result.failed or not result.data:
+ msg = result.RemoteFailMsg()
+ 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")
- raise errors.OpExecError("Can't add local storage to drbd")
+ raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
dev.children = new_lvs
cfg.Update(instance)
results = self.rpc.call_vg_list([pri_node, new_node])
for node in pri_node, new_node:
res = results[node]
- if res.failed or not res.data or my_vg not in res.data:
+ 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):
instance.hypervisor)
for node in nodenames:
info = nodeinfo[node]
- if info.failed or not info.data:
+ msg = info.RemoteFailMsg()
+ if msg:
raise errors.OpPrereqError("Cannot get current information"
- " from node '%s'" % node)
- vg_free = info.data.get('vg_free', None)
+ " from node %s:" % (node, msg))
+ vg_free = info.payload.get('vg_free', None)
if not isinstance(vg_free, int):
raise errors.OpPrereqError("Can't compute free disk space on"
" node %s" % node)
remote_info = self.rpc.call_instance_info(instance.primary_node,
instance.name,
instance.hypervisor)
- remote_info.Raise()
- remote_info = remote_info.data
+ msg = remote_info.RemoteFailMsg()
+ if msg:
+ raise errors.OpExecError("Error checking node %s: %s" %
+ (instance.primary_node, msg))
+ remote_info = remote_info.payload
if remote_info and "state" in remote_info:
remote_state = "up"
else:
if not utils.IsValidIP(nic_ip):
raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
+ 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")
+ elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
+ nic_dict['bridge'] = None
+ elif nic_link and nic_link.lower() == constants.VALUE_NONE:
+ nic_dict['link'] = None
+
if nic_op == constants.DDM_ADD:
- nic_bridge = nic_dict.get('bridge', None)
- if nic_bridge is None:
- nic_dict['bridge'] = self.cfg.GetDefBridge()
nic_mac = nic_dict.get('mac', None)
if nic_mac is None:
nic_dict['mac'] = constants.VALUE_AUTO
if self.op.nics:
args['nics'] = []
nic_override = dict(self.op.nics)
+ c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
for idx, nic in enumerate(self.instance.nics):
if idx in nic_override:
this_nic_override = nic_override[idx]
ip = this_nic_override['ip']
else:
ip = nic.ip
- if 'bridge' in this_nic_override:
- bridge = this_nic_override['bridge']
- else:
- bridge = nic.bridge
if 'mac' in this_nic_override:
mac = this_nic_override['mac']
else:
mac = nic.mac
- args['nics'].append((ip, bridge, mac))
+ if idx in self.nic_pnew:
+ nicparams = self.nic_pnew[idx]
+ else:
+ nicparams = objects.FillDict(c_nicparams, nic.nicparams)
+ mode = nicparams[constants.NIC_MODE]
+ link = nicparams[constants.NIC_LINK]
+ args['nics'].append((ip, mac, mode, link))
if constants.DDM_ADD in nic_override:
ip = nic_override[constants.DDM_ADD].get('ip', None)
- bridge = nic_override[constants.DDM_ADD]['bridge']
mac = nic_override[constants.DDM_ADD]['mac']
- args['nics'].append((ip, bridge, mac))
+ nicparams = self.nic_pnew[constants.DDM_ADD]
+ mode = nicparams[constants.NIC_MODE]
+ link = nicparams[constants.NIC_LINK]
+ args['nics'].append((ip, mac, mode, link))
elif constants.DDM_REMOVE in nic_override:
del args['nics'][-1]
nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
return env, nl, nl
+ def _GetUpdatedParams(self, old_params, update_dict,
+ default_values, parameter_types):
+ """Return the new params dict for the given params.
+
+ @type old_params: dict
+ @type 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
+ @type default_values: dict
+ @param default_values: default values for the filled parameters
+ @type parameter_types: dict
+ @param parameter_types: dict mapping target dict keys to types
+ in constants.ENFORCEABLE_TYPES
+ @rtype: (dict, dict)
+ @return: (new_parameters, filled_parameters)
+
+ """
+ params_copy = copy.deepcopy(old_params)
+ for key, val in update_dict.iteritems():
+ if val == constants.VALUE_DEFAULT:
+ try:
+ del params_copy[key]
+ except KeyError:
+ pass
+ else:
+ params_copy[key] = val
+ utils.ForceDictType(params_copy, parameter_types)
+ params_filled = objects.FillDict(default_values, params_copy)
+ return (params_copy, params_filled)
+
def CheckPrereq(self):
"""Check prerequisites.
# checking the new params on the primary/secondary nodes
instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
+ cluster = self.cluster = self.cfg.GetClusterInfo()
assert self.instance is not None, \
"Cannot retrieve locked instance %s" % self.op.instance_name
pnode = instance.primary_node
# hvparams processing
if self.op.hvparams:
- i_hvdict = copy.deepcopy(instance.hvparams)
- for key, val in self.op.hvparams.iteritems():
- if val == constants.VALUE_DEFAULT:
- try:
- del i_hvdict[key]
- except KeyError:
- pass
- else:
- i_hvdict[key] = val
- cluster = self.cfg.GetClusterInfo()
- utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
- hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
- i_hvdict)
+ i_hvdict, hv_new = self._GetUpdatedParams(
+ instance.hvparams, self.op.hvparams,
+ cluster.hvparams[instance.hypervisor],
+ constants.HVS_PARAMETER_TYPES)
# local check
hypervisor.GetHypervisor(
instance.hypervisor).CheckParameterSyntax(hv_new)
# beparams processing
if self.op.beparams:
- i_bedict = copy.deepcopy(instance.beparams)
- for key, val in self.op.beparams.iteritems():
- if val == constants.VALUE_DEFAULT:
- try:
- del i_bedict[key]
- except KeyError:
- pass
- else:
- i_bedict[key] = val
- cluster = self.cfg.GetClusterInfo()
- utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
- be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
- i_bedict)
+ i_bedict, be_new = self._GetUpdatedParams(
+ instance.beparams, self.op.beparams,
+ cluster.beparams[constants.PP_DEFAULT],
+ constants.BES_PARAMETER_TYPES)
self.be_new = be_new # the new actual values
self.be_inst = i_bedict # the new dict (without defaults)
else:
instance.hypervisor)
nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
instance.hypervisor)
- if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
+ pninfo = nodeinfo[pnode]
+ msg = pninfo.RemoteFailMsg()
+ if msg:
# Assume the primary node is unreachable and go ahead
- self.warn.append("Can't get info from primary node %s" % pnode)
+ self.warn.append("Can't get info from primary node %s: %s" %
+ (pnode, msg))
+ elif not isinstance(pninfo.payload.get('memory_free', None), int):
+ self.warn.append("Node data from primary node %s doesn't contain"
+ " free memory information" % pnode)
+ elif instance_info.RemoteFailMsg():
+ self.warn.append("Can't get instance runtime information: %s" %
+ instance_info.RemoteFailMsg())
else:
- if not instance_info.failed and instance_info.data:
- current_mem = instance_info.data['memory']
+ if instance_info.payload:
+ current_mem = int(instance_info.payload['memory'])
else:
# Assume instance not running
# (there is a slight race condition here, but it's not very probable,
# and we have no other way to check)
current_mem = 0
miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
- nodeinfo[pnode].data['memory_free'])
+ pninfo.payload['memory_free'])
if miss_mem > 0:
raise errors.OpPrereqError("This change will prevent the instance"
" from starting, due to %d MB of memory"
" missing on its primary node" % miss_mem)
if be_new[constants.BE_AUTO_BALANCE]:
- for node, nres in nodeinfo.iteritems():
+ for node, nres in nodeinfo.items():
if node not in instance.secondary_nodes:
continue
- if nres.failed or not isinstance(nres.data, dict):
- self.warn.append("Can't get info from secondary node %s" % node)
- elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
+ msg = nres.RemoteFailMsg()
+ if msg:
+ self.warn.append("Can't get info from secondary node %s: %s" %
+ (node, msg))
+ elif not isinstance(nres.payload.get('memory_free', None), int):
+ self.warn.append("Secondary node %s didn't return free"
+ " memory information" % node)
+ elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
self.warn.append("Not enough memory to failover instance to"
" secondary node %s" % node)
# NIC processing
+ self.nic_pnew = {}
+ self.nic_pinst = {}
for nic_op, nic_dict in self.op.nics:
if nic_op == constants.DDM_REMOVE:
if not instance.nics:
raise errors.OpPrereqError("Invalid NIC index %s, valid values"
" are 0 to %d" %
(nic_op, len(instance.nics)))
+ old_nic_params = instance.nics[nic_op].nicparams
+ old_nic_ip = instance.nics[nic_op].ip
+ else:
+ old_nic_params = {}
+ old_nic_ip = None
+
+ update_params_dict = dict([(key, nic_dict[key])
+ for key in constants.NICS_PARAMETERS
+ if key in nic_dict])
+
if 'bridge' in nic_dict:
- nic_bridge = nic_dict['bridge']
- if nic_bridge is None:
- raise errors.OpPrereqError('Cannot set the nic bridge to None')
- if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
- msg = ("Bridge '%s' doesn't exist on one of"
- " the instance nodes" % nic_bridge)
+ update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
+
+ new_nic_params, new_filled_nic_params = \
+ self._GetUpdatedParams(old_nic_params, update_params_dict,
+ cluster.nicparams[constants.PP_DEFAULT],
+ constants.NICS_PARAMETER_TYPES)
+ objects.NIC.CheckParameterSyntax(new_filled_nic_params)
+ self.nic_pinst[nic_op] = new_nic_params
+ self.nic_pnew[nic_op] = new_filled_nic_params
+ new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
+
+ if new_nic_mode == constants.NIC_MODE_BRIDGED:
+ nic_bridge = new_filled_nic_params[constants.NIC_LINK]
+ result = self.rpc.call_bridges_exist(pnode, [nic_bridge])
+ msg = result.RemoteFailMsg()
+ if msg:
+ msg = "Error checking bridges on node %s: %s" % (pnode, msg)
if self.force:
self.warn.append(msg)
else:
raise errors.OpPrereqError(msg)
+ if new_nic_mode == constants.NIC_MODE_ROUTED:
+ if 'ip' in nic_dict:
+ nic_ip = nic_dict['ip']
+ else:
+ nic_ip = old_nic_ip
+ if nic_ip is None:
+ raise errors.OpPrereqError('Cannot set the nic ip to None'
+ ' on a routed nic')
if 'mac' in nic_dict:
nic_mac = nic_dict['mac']
if nic_mac is None:
" an instance")
ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
ins_l = ins_l[pnode]
- if ins_l.failed or not isinstance(ins_l.data, list):
- raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
- if instance.name in ins_l.data:
+ msg = ins_l.RemoteFailMsg()
+ if msg:
+ raise errors.OpPrereqError("Can't contact node %s: %s" %
+ (pnode, msg))
+ if instance.name in ins_l.payload:
raise errors.OpPrereqError("Instance is running, can't remove"
" disks.")
result = []
instance = self.instance
+ cluster = self.cluster
# disk changes
for disk_op, disk_dict in self.op.disks:
if disk_op == constants.DDM_REMOVE:
elif nic_op == constants.DDM_ADD:
# mac and bridge should be set, by now
mac = nic_dict['mac']
- bridge = nic_dict['bridge']
- new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
- bridge=bridge)
+ ip = nic_dict.get('ip', None)
+ nicparams = self.nic_pinst[constants.DDM_ADD]
+ new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
instance.nics.append(new_nic)
result.append(("nic.%d" % (len(instance.nics) - 1),
- "add:mac=%s,ip=%s,bridge=%s" %
- (new_nic.mac, new_nic.ip, new_nic.bridge)))
+ "add:mac=%s,ip=%s,mode=%s,link=%s" %
+ (new_nic.mac, new_nic.ip,
+ self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
+ self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
+ )))
else:
- # change a given nic
- for key in 'mac', 'ip', 'bridge':
+ for key in 'mac', 'ip':
if key in nic_dict:
setattr(instance.nics[nic_op], key, nic_dict[key])
- result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
+ if nic_op in self.nic_pnew:
+ instance.nics[nic_op].nicparams = self.nic_pnew[nic_op]
+ for key, val in nic_dict.iteritems():
+ result.append(("nic.%s/%d" % (key, nic_op), val))
# hvparams changes
if self.op.hvparams:
rpcresult = self.rpc.call_export_list(self.nodes)
result = {}
for node in rpcresult:
- if rpcresult[node].failed:
+ if rpcresult[node].RemoteFailMsg():
result[node] = False
else:
- result[node] = rpcresult[node].data
+ result[node] = rpcresult[node].payload
return result
try:
for disk in instance.disks:
- # new_dev_name will be a snapshot of an lvm leaf of the one we passed
- new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
- if new_dev_name.failed or not new_dev_name.data:
- self.LogWarning("Could not snapshot block device %s on node %s",
- disk.logical_id[1], src_node)
+ # 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()
+ if msg:
+ self.LogWarning("Could not snapshot block device %s on node %s: %s",
+ disk.logical_id[1], src_node, msg)
snap_disks.append(False)
else:
+ disk_id = (vgname, result.payload)
new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
- logical_id=(vgname, new_dev_name.data),
- physical_id=(vgname, new_dev_name.data),
+ logical_id=disk_id, physical_id=disk_id,
iv_name=disk.iv_name)
snap_disks.append(new_dev)
finally:
if self.op.shutdown and instance.admin_up:
- result = self.rpc.call_instance_start(src_node, instance)
+ result = self.rpc.call_instance_start(src_node, instance, None, None)
msg = result.RemoteFailMsg()
if msg:
_ShutdownInstanceDisks(self, instance)
if dev:
result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
instance, cluster_name, idx)
- if result.failed or not result.data:
+ msg = result.RemoteFailMsg()
+ if msg:
self.LogWarning("Could not export block device %s from node %s to"
- " node %s", dev.logical_id[1], src_node,
- dst_node.name)
+ " node %s: %s", dev.logical_id[1], src_node,
+ dst_node.name, msg)
msg = self.rpc.call_blockdev_remove(src_node, dev).RemoteFailMsg()
if msg:
self.LogWarning("Could not remove snapshot block device %s from node"
" %s: %s", dev.logical_id[1], src_node, msg)
result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
- if result.failed or not result.data:
- self.LogWarning("Could not finalize export for instance %s on node %s",
- instance.name, dst_node.name)
+ msg = result.RemoteFailMsg()
+ if msg:
+ self.LogWarning("Could not finalize export for instance %s"
+ " on node %s: %s", instance.name, dst_node.name, msg)
nodelist = self.cfg.GetNodeList()
nodelist.remove(dst_node.name)
# on one-node clusters nodelist will be empty after the removal
# if we proceed the backup would be removed because OpQueryExports
# substitutes an empty list with the full cluster node list.
+ iname = instance.name
if nodelist:
exportlist = self.rpc.call_export_list(nodelist)
for node in exportlist:
- if exportlist[node].failed:
+ if exportlist[node].RemoteFailMsg():
continue
- if instance.name in exportlist[node].data:
- if not self.rpc.call_export_remove(node, instance.name):
+ if iname in exportlist[node].payload:
+ msg = self.rpc.call_export_remove(node, iname).RemoteFailMsg()
+ if msg:
self.LogWarning("Could not remove older export for instance %s"
- " on node %s", instance.name, node)
+ " on node %s: %s", iname, node, msg)
class LURemoveExport(NoHooksLU):
fqdn_warn = True
instance_name = self.op.instance_name
- exportlist = self.rpc.call_export_list(self.acquired_locks[
- locking.LEVEL_NODE])
+ locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
+ exportlist = self.rpc.call_export_list(locked_nodes)
found = False
for node in exportlist:
- if exportlist[node].failed:
- self.LogWarning("Failed to query node %s, continuing" % node)
+ msg = exportlist[node].RemoteFailMsg()
+ if msg:
+ self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
continue
- if instance_name in exportlist[node].data:
+ if instance_name in exportlist[node].payload:
found = True
result = self.rpc.call_export_remove(node, instance_name)
- if result.failed or not result.data:
+ msg = result.RemoteFailMsg()
+ if msg:
logging.error("Could not remove export for instance %s"
- " on node %s", instance_name, node)
+ " on node %s: %s", instance_name, node, msg)
if fqdn_warn and not found:
feedback_fn("Export not found. If trying to remove an export belonging"
}
if not ninfo.offline:
- nresult.Raise()
- if not isinstance(nresult.data, dict):
- raise errors.OpExecError("Can't get data for node %s" % nname)
- remote_info = nresult.data
+ 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))
+ remote_info = nresult.payload
for attr in ['memory_total', 'memory_free', 'memory_dom0',
'vg_size', 'vg_free', 'cpu_total']:
if attr not in remote_info:
raise errors.OpExecError("Node '%s' didn't return attribute"
" '%s'" % (nname, attr))
- try:
- remote_info[attr] = int(remote_info[attr])
- except ValueError, err:
+ if not isinstance(remote_info[attr], int):
raise errors.OpExecError("Node '%s' returned invalid value"
- " for '%s': %s" % (nname, attr, err))
+ " for '%s': %s" %
+ (nname, attr, remote_info[attr]))
# compute memory used by primary instances
i_p_mem = i_p_up_mem = 0
for iinfo, beinfo in i_list:
if iinfo.primary_node == nname:
i_p_mem += beinfo[constants.BE_MEMORY]
- if iinfo.name not in node_iinfo[nname].data:
+ if iinfo.name not in node_iinfo[nname].payload:
i_used_mem = 0
else:
- i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
+ i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
remote_info['memory_free'] -= max(0, i_mem_diff)
# instance data
instance_data = {}
for iinfo, beinfo in i_list:
- nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
- for n in iinfo.nics]
+ nic_data = []
+ for nic in iinfo.nics:
+ filled_params = objects.FillDict(
+ cluster_info.nicparams[constants.PP_DEFAULT],
+ nic.nicparams)
+ nic_dict = {"mac": nic.mac,
+ "ip": nic.ip,
+ "mode": filled_params[constants.NIC_MODE],
+ "link": filled_params[constants.NIC_LINK],
+ }
+ if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
+ nic_dict["bridge"] = filled_params[constants.NIC_LINK]
+ nic_data.append(nic_dict)
pir = {
"tags": list(iinfo.GetTags()),
"admin_up": iinfo.admin_up,
"disk_template": iinfo.disk_template,
"hypervisor": iinfo.hypervisor,
}
+ pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
+ pir["disks"])
instance_data[iinfo.name] = pir
data["instances"] = instance_data
projects / ganeti-local / blobdiff