#
#
-# Copyright (C) 2006, 2007 Google Inc.
+# Copyright (C) 2006, 2007, 2008 Google Inc.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
added by the hooks runner. If the LU doesn't define any
environment, an empty dict (and not None) should be returned.
- As for the node lists, the master should not be included in the
- them, as it will be added by the hooks runner in case this LU
- requires a cluster to run on (otherwise we don't have a node
- list). No nodes should be returned as an empty list (and not
- None).
+ No nodes should be returned as an empty list (and not None).
Note that if the HPATH for a LU class is None, this function will
not be called.
"""
raise NotImplementedError
+ def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
+ """Notify the LU about the results of its hooks.
+
+ This method is called every time a hooks phase is executed, and notifies
+ the Logical Unit about the hooks' result. The LU can then use it to alter
+ its result based on the hooks. By default the method does nothing and the
+ previous result is passed back unchanged but any LU can define it if it
+ wants to use the local cluster hook-scripts somehow.
+
+ Args:
+ phase: the hooks phase that has just been run
+ hooks_results: the results of the multi-node hooks rpc call
+ feedback_fn: function to send feedback back to the caller
+ lu_result: the previous result this LU had, or None in the PRE phase.
+
+ """
+ return lu_result
+
class NoHooksLU(LogicalUnit):
"""Simple LU which runs no hooks.
HPATH = None
HTYPE = None
- def BuildHooksEnv(self):
- """Build hooks env.
-
- This is a no-op, since we don't run hooks.
-
- """
- return {}, [], []
-
def _AddHostToEtcHosts(hostname):
"""Wrapper around utils.SetEtcHostsEntry.
rpc.call_node_leave_cluster(master)
-class LUVerifyCluster(NoHooksLU):
+class LUVerifyCluster(LogicalUnit):
"""Verifies the cluster status.
"""
+ HPATH = "cluster-verify"
+ HTYPE = constants.HTYPE_CLUSTER
_OP_REQP = ["skip_checks"]
def _VerifyNode(self, node, file_list, local_cksum, vglist, node_result,
if 'nodelist' not in node_result:
bad = True
- feedback_fn(" - ERROR: node hasn't returned node connectivity data")
+ feedback_fn(" - ERROR: node hasn't returned node ssh connectivity data")
else:
if node_result['nodelist']:
bad = True
for node in node_result['nodelist']:
- feedback_fn(" - ERROR: communication with node '%s': %s" %
+ feedback_fn(" - ERROR: ssh communication with node '%s': %s" %
(node, node_result['nodelist'][node]))
+ if 'node-net-test' not in node_result:
+ bad = True
+ feedback_fn(" - ERROR: node hasn't returned node tcp connectivity data")
+ else:
+ if node_result['node-net-test']:
+ bad = True
+ nlist = utils.NiceSort(node_result['node-net-test'].keys())
+ for node in nlist:
+ feedback_fn(" - ERROR: tcp communication with node '%s': %s" %
+ (node, node_result['node-net-test'][node]))
+
hyp_result = node_result.get('hypervisor', None)
if hyp_result is not None:
feedback_fn(" - ERROR: hypervisor verify failure: '%s'" % hyp_result)
for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
needed_mem = 0
for instance in instances:
- needed_mem += instance_cfg[instance].memory
+ if instance_cfg[instance].auto_balance:
+ needed_mem += instance_cfg[instance].memory
if nodeinfo['mfree'] < needed_mem:
feedback_fn(" - ERROR: not enough memory on node %s to accomodate"
" failovers should node %s fail" % (node, prinode))
if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
raise errors.OpPrereqError("Invalid checks to be skipped specified")
+ def BuildHooksEnv(self):
+ """Build hooks env.
+
+ Cluster-Verify hooks just rone in the post phase and their failure makes
+ the output be logged in the verify output and the verification to fail.
+
+ """
+ all_nodes = self.cfg.GetNodeList()
+ tags = self.cfg.GetClusterInfo().GetTags()
+ # TODO: populate the environment with useful information for verify hooks
+ env = {
+ "CLUSTER_TAGS": " ".join(tags),
+ }
+ return env, [], all_nodes
+
def Exec(self, feedback_fn):
"""Verify integrity of cluster, performing various test on nodes.
vg_name = self.cfg.GetVGName()
nodelist = utils.NiceSort(self.cfg.GetNodeList())
+ nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
instancelist = utils.NiceSort(self.cfg.GetInstanceList())
i_non_redundant = [] # Non redundant instances
+ i_non_a_balanced = [] # Non auto-balanced instances
node_volume = {}
node_instance = {}
node_info = {}
'filelist': file_names,
'nodelist': nodelist,
'hypervisor': None,
+ 'node-net-test': [(node.name, node.primary_ip, node.secondary_ip)
+ for node in nodeinfo]
}
all_nvinfo = rpc.call_node_verify(nodelist, node_verify_param)
all_rversion = rpc.call_version(nodelist)
all_ninfo = rpc.call_node_info(nodelist, self.cfg.GetVGName())
+ incomplete_nodeinfo = False
+
for node in nodelist:
feedback_fn("* Verifying node %s" % node)
result = self._VerifyNode(node, file_names, local_checksums,
elif not isinstance(volumeinfo, dict):
feedback_fn(" - ERROR: connection to %s failed" % (node,))
bad = True
+ incomplete_nodeinfo = True
continue
else:
node_volume[node] = volumeinfo
if type(nodeinstance) != list:
feedback_fn(" - ERROR: connection to %s failed" % (node,))
bad = True
+ incomplete_nodeinfo = True
continue
node_instance[node] = nodeinstance
if not isinstance(nodeinfo, dict):
feedback_fn(" - ERROR: connection to %s failed" % (node,))
bad = True
+ incomplete_nodeinfo = True
continue
try:
# secondary.
"sinst-by-pnode": {},
}
- except ValueError:
+ except (ValueError, TypeError):
feedback_fn(" - ERROR: invalid value returned from node %s" % (node,))
bad = True
+ incomplete_nodeinfo = True
continue
node_vol_should = {}
feedback_fn(" - WARNING: multiple secondaries for instance %s"
% instance)
+ if not inst_config.auto_balance:
+ i_non_a_balanced.append(instance)
+
for snode in inst_config.secondary_nodes:
if snode in node_info:
node_info[snode]['sinst'].append(instance)
feedback_fn)
bad = bad or result
- if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
+ if (constants.VERIFY_NPLUSONE_MEM not in self.skip_set and
+ not incomplete_nodeinfo):
feedback_fn("* Verifying N+1 Memory redundancy")
result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
bad = bad or result
feedback_fn(" - NOTICE: %d non-redundant instance(s) found."
% len(i_non_redundant))
+ if i_non_a_balanced:
+ feedback_fn(" - NOTICE: %d non-auto-balanced instance(s) found."
+ % len(i_non_a_balanced))
+
return int(bad)
+ def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
+ """Analize the post-hooks' result, handle it, and send some
+ nicely-formatted feedback back to the user.
+
+ Args:
+ phase: the hooks phase that has just been run
+ hooks_results: the results of the multi-node hooks rpc call
+ feedback_fn: function to send feedback back to the caller
+ lu_result: previous Exec result
+
+ """
+ # We only really run POST phase hooks, and are only interested in their results
+ if phase == constants.HOOKS_PHASE_POST:
+ # Used to change hooks' output to proper indentation
+ indent_re = re.compile('^', re.M)
+ feedback_fn("* Hooks Results")
+ if not hooks_results:
+ feedback_fn(" - ERROR: general communication failure")
+ lu_result = 1
+ else:
+ for node_name in hooks_results:
+ show_node_header = True
+ res = hooks_results[node_name]
+ if res is False or not isinstance(res, list):
+ feedback_fn(" Communication failure")
+ lu_result = 1
+ continue
+ for script, hkr, output in res:
+ if hkr == constants.HKR_FAIL:
+ # The node header is only shown once, if there are
+ # failing hooks on that node
+ if show_node_header:
+ feedback_fn(" Node %s:" % node_name)
+ show_node_header = False
+ feedback_fn(" ERROR: Script %s failed, output:" % script)
+ output = indent_re.sub(' ', output)
+ feedback_fn("%s" % output)
+ lu_result = 1
+
+ return lu_result
+
class LUVerifyDisks(NoHooksLU):
"""Verifies the cluster disks status.
raise errors.OpPrereqError("Neither the name nor the IP address of the"
" cluster has changed")
if new_ip != old_ip:
- result = utils.RunCmd(["fping", "-q", new_ip])
- if not result.failed:
+ if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
raise errors.OpPrereqError("The given cluster IP address (%s) is"
" reachable on the network. Aborting." %
new_ip)
"""Logical unit for OS diagnose/query.
"""
- _OP_REQP = []
+ _OP_REQP = ["output_fields", "names"]
def CheckPrereq(self):
"""Check prerequisites.
This always succeeds, since this is a pure query LU.
"""
- return
+ if self.op.names:
+ raise errors.OpPrereqError("Selective OS query not supported")
+
+ self.dynamic_fields = frozenset(["name", "valid", "node_status"])
+ _CheckOutputFields(static=[],
+ dynamic=self.dynamic_fields,
+ selected=self.op.output_fields)
+
+ @staticmethod
+ def _DiagnoseByOS(node_list, rlist):
+ """Remaps a per-node return list into an a per-os per-node dictionary
+
+ Args:
+ node_list: a list with the names of all nodes
+ rlist: a map with node names as keys and OS objects as values
+
+ Returns:
+ map: a map with osnames as keys and as value another map, with
+ nodes as
+ keys and list of OS objects as values
+ e.g. {"debian-etch": {"node1": [<object>,...],
+ "node2": [<object>,]}
+ }
+
+ """
+ all_os = {}
+ for node_name, nr in rlist.iteritems():
+ if not nr:
+ continue
+ for os in nr:
+ if os.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.name] = {}
+ for nname in node_list:
+ all_os[os.name][nname] = []
+ all_os[os.name][node_name].append(os)
+ return all_os
def Exec(self, feedback_fn):
"""Compute the list of OSes.
node_data = rpc.call_os_diagnose(node_list)
if node_data == False:
raise errors.OpExecError("Can't gather the list of OSes")
- return node_data
+ pol = self._DiagnoseByOS(node_list, node_data)
+ output = []
+ for os_name, os_data in pol.iteritems():
+ 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()])
+ elif field == "node_status":
+ val = {}
+ for node_name, nos_list in os_data.iteritems():
+ val[node_name] = [(v.status, v.path) for v in nos_list]
+ else:
+ raise errors.ParameterError(field)
+ row.append(val)
+ output.append(row)
+
+ return output
class LURemoveNode(LogicalUnit):
This checks that the fields required are valid output fields.
"""
- self.dynamic_fields = frozenset(["dtotal", "dfree",
- "mtotal", "mnode", "mfree",
- "bootid"])
+ self.dynamic_fields = frozenset([
+ "dtotal", "dfree",
+ "mtotal", "mnode", "mfree",
+ "bootid",
+ "ctotal",
+ ])
_CheckOutputFields(static=["name", "pinst_cnt", "sinst_cnt",
"pinst_list", "sinst_list",
- "pip", "sip"],
+ "pip", "sip", "tags"],
dynamic=self.dynamic_fields,
selected=self.op.output_fields)
"mfree": utils.TryConvert(int, nodeinfo['memory_free']),
"dtotal": utils.TryConvert(int, nodeinfo['vg_size']),
"dfree": utils.TryConvert(int, nodeinfo['vg_free']),
+ "ctotal": utils.TryConvert(int, nodeinfo['cpu_total']),
"bootid": nodeinfo['bootid'],
}
else:
val = node.primary_ip
elif field == "sip":
val = node.secondary_ip
+ elif field == "tags":
+ val = list(node.GetTags())
elif field in self.dynamic_fields:
val = live_data[node.name].get(field, None)
else:
if not utils.IsValidIP(secondary_ip):
raise errors.OpPrereqError("Invalid secondary IP given")
self.op.secondary_ip = secondary_ip
+
node_list = cfg.GetNodeList()
- if node in node_list:
- raise errors.OpPrereqError("Node %s is already in the configuration"
- % node)
+ if not self.op.readd and node in node_list:
+ raise errors.OpPrereqError("Node %s is already in the configuration" %
+ node)
+ elif self.op.readd and node not in node_list:
+ raise errors.OpPrereqError("Node %s is not in the configuration" % node)
for existing_node_name in node_list:
existing_node = cfg.GetNodeInfo(existing_node_name)
+
+ if self.op.readd and node == existing_node_name:
+ if (existing_node.primary_ip != primary_ip or
+ existing_node.secondary_ip != secondary_ip):
+ raise errors.OpPrereqError("Readded node doesn't have the same IP"
+ " address configuration as before")
+ continue
+
if (existing_node.primary_ip == primary_ip or
existing_node.secondary_ip == primary_ip or
existing_node.primary_ip == secondary_ip or
# Distribute updated /etc/hosts and known_hosts to all nodes,
# including the node just added
myself = self.cfg.GetNodeInfo(self.sstore.GetMasterNode())
- dist_nodes = self.cfg.GetNodeList() + [node]
+ 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)
if not ssh.CopyFileToNode(node, fname):
logger.Error("could not copy file %s to node %s" % (fname, node))
- logger.Info("adding node %s to cluster.conf" % node)
- self.cfg.AddNode(new_node)
+ if not self.op.readd:
+ logger.Info("adding node %s to cluster.conf" % node)
+ self.cfg.AddNode(new_node)
class LUMasterFailover(LogicalUnit):
"export_version": constants.EXPORT_VERSION,
"master": self.sstore.GetMasterNode(),
"architecture": (platform.architecture()[0], platform.machine()),
+ "hypervisor_type": self.sstore.GetHypervisorType(),
}
return result
"""
nodeinfo = rpc.call_node_info([node], cfg.GetVGName())
- if not nodeinfo or not isinstance(nodeinfo, dict):
+ if not (nodeinfo and isinstance(nodeinfo, dict) and
+ node in nodeinfo and isinstance(nodeinfo[node], dict)):
raise errors.OpPrereqError("Could not contact node %s for resource"
" information" % (node,))
new_name)
if not getattr(self.op, "ignore_ip", False):
- command = ["fping", "-q", name_info.ip]
- result = utils.RunCmd(command)
- if not result.failed:
+ if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
raise errors.OpPrereqError("IP %s of instance %s already in use" %
(name_info.ip, new_name))
try:
if not rpc.call_instance_run_rename(inst.primary_node, inst, old_name,
"sda", "sdb"):
- msg = ("Could run OS rename script for instance %s on node %s (but the"
- " instance has been renamed in Ganeti)" %
+ msg = ("Could not run OS rename script for instance %s on node %s"
+ " (but the instance has been renamed in Ganeti)" %
(inst.name, inst.primary_node))
logger.Error(msg)
finally:
"""
HPATH = "instance-remove"
HTYPE = constants.HTYPE_INSTANCE
- _OP_REQP = ["instance_name"]
+ _OP_REQP = ["instance_name", "ignore_failures"]
def BuildHooksEnv(self):
"""Build hooks env.
_CheckOutputFields(static=["name", "os", "pnode", "snodes",
"admin_state", "admin_ram",
"disk_template", "ip", "mac", "bridge",
- "sda_size", "sdb_size", "vcpus"],
+ "sda_size", "sdb_size", "vcpus", "tags",
+ "auto_balance",
+ "network_port", "kernel_path", "initrd_path",
+ "hvm_boot_order", "hvm_acpi", "hvm_pae",
+ "hvm_cdrom_image_path", "hvm_nic_type",
+ "hvm_disk_type", "vnc_bind_address"],
dynamic=self.dynamic_fields,
selected=self.op.output_fields)
val = disk.size
elif field == "vcpus":
val = instance.vcpus
+ elif field == "tags":
+ val = list(instance.GetTags())
+ elif field == "auto_balance":
+ val = instance.auto_balance
+ elif field in ("network_port", "kernel_path", "initrd_path",
+ "hvm_boot_order", "hvm_acpi", "hvm_pae",
+ "hvm_cdrom_image_path", "hvm_nic_type",
+ "hvm_disk_type", "vnc_bind_address"):
+ val = getattr(instance, field, None)
+ if val is None:
+ if field in ("hvm_nic_type", "hvm_disk_type",
+ "kernel_path", "initrd_path"):
+ val = "default"
+ else:
+ val = "-"
else:
raise errors.ParameterError(field)
iout.append(val)
instance.primary_node = target_node
# distribute new instance config to the other nodes
- self.cfg.AddInstance(instance)
+ self.cfg.Update(instance)
# Only start the instance if it's marked as up
if instance.status == "up":
(instance.name, target_node))
+class LUMigrateInstance(LogicalUnit):
+ """Migrate an instance.
+
+ This is migration without shutting down, compared to the failover,
+ which is done with shutdown.
+
+ """
+ HPATH = "instance-migrate"
+ HTYPE = constants.HTYPE_INSTANCE
+ _OP_REQP = ["instance_name", "live", "cleanup"]
+
+ def BuildHooksEnv(self):
+ """Build hooks env.
+
+ This runs on master, primary and secondary nodes of the instance.
+
+ """
+ env = _BuildInstanceHookEnvByObject(self.instance)
+ nl = [self.sstore.GetMasterNode()] + list(self.instance.secondary_nodes)
+ return env, nl, nl
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This checks that the instance is in the cluster.
+
+ """
+ instance = self.cfg.GetInstanceInfo(
+ self.cfg.ExpandInstanceName(self.op.instance_name))
+ if instance is None:
+ raise errors.OpPrereqError("Instance '%s' not known" %
+ self.op.instance_name)
+
+ if instance.disk_template != constants.DT_DRBD8:
+ raise errors.OpPrereqError("Instance's disk layout is not"
+ " drbd8, cannot migrate.")
+
+ secondary_nodes = instance.secondary_nodes
+ if not secondary_nodes:
+ raise errors.ProgrammerError("no secondary node but using "
+ "drbd8 disk template")
+
+ target_node = secondary_nodes[0]
+ # check memory requirements on the secondary node
+ _CheckNodeFreeMemory(self.cfg, target_node, "migrating instance %s" %
+ instance.name, instance.memory)
+
+ # check bridge existance
+ brlist = [nic.bridge for nic in instance.nics]
+ if not rpc.call_bridges_exist(target_node, brlist):
+ raise errors.OpPrereqError("One or more target bridges %s does not"
+ " exist on destination node '%s'" %
+ (brlist, target_node))
+
+ if not self.op.cleanup:
+ migratable = rpc.call_instance_migratable(instance.primary_node,
+ instance)
+ if not migratable:
+ raise errors.OpPrereqError("Can't contact node '%s'" %
+ instance.primary_node)
+ if not migratable[0]:
+ raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
+ migratable[1])
+
+ self.instance = instance
+
+ def _WaitUntilSync(self):
+ """Poll with custom rpc for disk sync.
+
+ This uses our own step-based rpc call.
+
+ """
+ self.feedback_fn("* wait until resync is done")
+ all_done = False
+ while not all_done:
+ all_done = True
+ result = rpc.call_drbd_reconfig_net(self.all_nodes, self.instance.name,
+ self.instance.disks,
+ self.nodes_ip, False,
+ constants.DRBD_RECONF_RPC_WFSYNC)
+ min_percent = 100
+ for node in self.all_nodes:
+ if not result[node] or not result[node][0]:
+ raise errors.OpExecError("Cannot resync disks on node %s" % (node,))
+ node_done, node_percent = result[node][1]
+ all_done = all_done and node_done
+ if node_percent is not None:
+ min_percent = min(min_percent, node_percent)
+ if not all_done:
+ if min_percent < 100:
+ self.feedback_fn(" - progress: %.1f%%" % min_percent)
+ time.sleep(2)
+
+ def _EnsureSecondary(self, node):
+ """Demote a node to secondary.
+
+ """
+ self.feedback_fn("* switching node %s to secondary mode" % node)
+ result = rpc.call_drbd_reconfig_net([node], self.instance.name,
+ self.instance.disks,
+ self.nodes_ip, False,
+ constants.DRBD_RECONF_RPC_SECONDARY)
+ if not result[node] or not result[node][0]:
+ raise errors.OpExecError("Cannot change disk to secondary on node %s,"
+ " error %s" %
+ (node, result[node][1]))
+
+ def _GoStandalone(self):
+ """Disconnect from the network.
+
+ """
+ self.feedback_fn("* changing into standalone mode")
+ result = rpc.call_drbd_reconfig_net(self.all_nodes, self.instance.name,
+ self.instance.disks,
+ self.nodes_ip, True,
+ constants.DRBD_RECONF_RPC_DISCONNECT)
+ for node in self.all_nodes:
+ if not result[node] or not result[node][0]:
+ raise errors.OpExecError("Cannot disconnect disks node %s,"
+ " error %s" % (node, result[node][1]))
+
+ def _GoReconnect(self, multimaster):
+ """Reconnect to the network.
+
+ """
+ if multimaster:
+ msg = "dual-master"
+ else:
+ msg = "single-master"
+ self.feedback_fn("* changing disks into %s mode" % msg)
+ result = rpc.call_drbd_reconfig_net(self.all_nodes, self.instance.name,
+ self.instance.disks,
+ self.nodes_ip,
+ multimaster,
+ constants.DRBD_RECONF_RPC_RECONNECT)
+ for node in self.all_nodes:
+ if not result[node] or not result[node][0]:
+ raise errors.OpExecError("Cannot change disks config on node %s,"
+ " error %s" % (node, result[node][1]))
+
+ def _IdentifyDisks(self):
+ """Start the migration RPC sequence.
+
+ """
+ self.feedback_fn("* identifying disks")
+ result = rpc.call_drbd_reconfig_net(self.all_nodes,
+ self.instance.name,
+ self.instance.disks,
+ self.nodes_ip, True,
+ constants.DRBD_RECONF_RPC_INIT)
+ for node in self.all_nodes:
+ if not result[node] or not result[node][0]:
+ raise errors.OpExecError("Cannot identify disks node %s,"
+ " error %s" % (node, result[node][1]))
+
+ def _ExecCleanup(self):
+ """Try to cleanup after a failed migration.
+
+ The cleanup is done by:
+ - check that the instance is running only on one node
+ (and update the config if needed)
+ - change disks on its secondary node to secondary
+ - wait until disks are fully synchronized
+ - disconnect from the network
+ - change disks into single-master mode
+ - wait again until disks are fully synchronized
+
+ """
+ instance = self.instance
+ target_node = self.target_node
+ source_node = self.source_node
+
+ # check running on only one node
+ self.feedback_fn("* checking where the instance actually runs"
+ " (if this hangs, the hypervisor might be in"
+ " a bad state)")
+ ins_l = rpc.call_instance_list(self.all_nodes)
+ for node in self.all_nodes:
+ if not type(ins_l[node]) is list:
+ raise errors.OpExecError("Can't contact node '%s'" % node)
+
+ runningon_source = instance.name in ins_l[source_node]
+ runningon_target = instance.name in ins_l[target_node]
+
+ if runningon_source and runningon_target:
+ raise errors.OpExecError("Instance seems to be running on two nodes,"
+ " or the hypervisor is confused. You will have"
+ " to ensure manually that it runs only on one"
+ " and restart this operation.")
+
+ if not (runningon_source or runningon_target):
+ raise errors.OpExecError("Instance does not seem to be running at all."
+ " In this case, it's safer to repair by"
+ " running 'gnt-instance stop' to ensure disk"
+ " shutdown, and then restarting it.")
+
+ if runningon_target:
+ # the migration has actually succeeded, we need to update the config
+ self.feedback_fn("* instance running on secondary node (%s),"
+ " updating config" % target_node)
+ instance.primary_node = target_node
+ self.cfg.Update(instance)
+ demoted_node = source_node
+ else:
+ self.feedback_fn("* instance confirmed to be running on its"
+ " primary node (%s)" % source_node)
+ demoted_node = target_node
+
+ self._IdentifyDisks()
+
+ self._EnsureSecondary(demoted_node)
+ self._WaitUntilSync()
+ self._GoStandalone()
+ self._GoReconnect(False)
+ self._WaitUntilSync()
+
+ self.feedback_fn("* done")
+
+ def _ExecMigration(self):
+ """Migrate an instance.
+
+ The migrate is done by:
+ - change the disks into dual-master mode
+ - wait until disks are fully synchronized again
+ - migrate the instance
+ - change disks on the new secondary node (the old primary) to secondary
+ - wait until disks are fully synchronized
+ - change disks into single-master mode
+
+ """
+ instance = self.instance
+ target_node = self.target_node
+ source_node = self.source_node
+
+ self.feedback_fn("* checking disk consistency between source and target")
+ for dev in instance.disks:
+ if not _CheckDiskConsistency(self.cfg, dev, target_node, False):
+ raise errors.OpExecError("Disk %s is degraded or not fully"
+ " synchronized on target node,"
+ " aborting migrate." % dev.iv_name)
+
+ self._IdentifyDisks()
+
+ self._EnsureSecondary(target_node)
+ self._GoStandalone()
+ self._GoReconnect(True)
+ self._WaitUntilSync()
+
+ self.feedback_fn("* migrating instance to %s" % target_node)
+ time.sleep(10)
+ result = rpc.call_instance_migrate(source_node, instance,
+ self.nodes_ip[target_node],
+ self.op.live)
+ if not result or not result[0]:
+ logger.Error("Instance migration failed, trying to revert disk status")
+ try:
+ self._EnsureSecondary(target_node)
+ self._GoStandalone()
+ self._GoReconnect(False)
+ self._WaitUntilSync()
+ except errors.OpExecError, err:
+ logger.Error("Can't reconnect the drives: error '%s'\n"
+ "Please look and recover the instance status" % str(err))
+
+ raise errors.OpExecError("Could not migrate instance %s: %s" %
+ (instance.name, result[1]))
+ time.sleep(10)
+
+ instance.primary_node = target_node
+ # distribute new instance config to the other nodes
+ self.cfg.Update(instance)
+
+ self._EnsureSecondary(source_node)
+ self._WaitUntilSync()
+ self._GoStandalone()
+ self._GoReconnect(False)
+ self._WaitUntilSync()
+
+ self.feedback_fn("* done")
+
+ def Exec(self, feedback_fn):
+ """Perform the migration.
+
+ """
+ self.feedback_fn = feedback_fn
+
+ self.source_node = self.instance.primary_node
+ self.target_node = self.instance.secondary_nodes[0]
+ self.all_nodes = [self.source_node, self.target_node]
+ self.nodes_ip = {
+ 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:
+ return self._ExecCleanup()
+ else:
+ return self._ExecMigration()
+
+
def _CreateBlockDevOnPrimary(cfg, node, instance, device, info):
"""Create a tree of block devices on the primary node.
HTYPE = constants.HTYPE_INSTANCE
_OP_REQP = ["instance_name", "mem_size", "disk_size",
"disk_template", "swap_size", "mode", "start", "vcpus",
- "wait_for_sync", "ip_check", "mac"]
+ "wait_for_sync", "ip_check", "mac", "auto_balance"]
def _RunAllocator(self):
"""Run the allocator based on input opcode.
nics = [{"mac": self.op.mac, "ip": getattr(self.op, "ip", None),
"bridge": self.op.bridge}]
ial = IAllocator(self.cfg, self.sstore,
+ mode=constants.IALLOCATOR_MODE_ALLOC,
name=self.op.instance_name,
disk_template=self.op.disk_template,
tags=[],
mem_size=self.op.mem_size,
disks=disks,
nics=nics,
- mode=constants.IALLOCATOR_MODE_ALLOC)
+ )
ial.Run(self.op.iallocator)
"""
# set optional parameters to none if they don't exist
for attr in ["kernel_path", "initrd_path", "hvm_boot_order", "pnode",
- "iallocator"]:
+ "iallocator", "hvm_acpi", "hvm_pae", "hvm_cdrom_image_path",
+ "hvm_nic_type", "hvm_disk_type", "vnc_bind_address"]:
if not hasattr(self.op, attr):
setattr(self.op, attr, None)
info = nodeinfo.get(node, None)
if not info:
raise errors.OpPrereqError("Cannot get current information"
- " from node '%s'" % nodeinfo)
+ " from node '%s'" % node)
vg_free = info.get('vg_free', None)
if not isinstance(vg_free, int):
raise errors.OpPrereqError("Can't compute free disk space on"
" destination node '%s'" %
(self.op.bridge, pnode.name))
+ # memory check on primary node
+ if self.op.start:
+ _CheckNodeFreeMemory(self.cfg, self.pnode.name,
+ "creating instance %s" % self.op.instance_name,
+ self.op.mem_size)
+
+ # hvm_cdrom_image_path verification
+ if self.op.hvm_cdrom_image_path is not None:
+ if not os.path.isabs(self.op.hvm_cdrom_image_path):
+ raise errors.OpPrereqError("The path to the HVM CDROM image must"
+ " be an absolute path or None, not %s" %
+ self.op.hvm_cdrom_image_path)
+ if not os.path.isfile(self.op.hvm_cdrom_image_path):
+ raise errors.OpPrereqError("The HVM CDROM image must either be a"
+ " regular file or a symlink pointing to"
+ " an existing regular file, not %s" %
+ self.op.hvm_cdrom_image_path)
+
+ # vnc_bind_address verification
+ if self.op.vnc_bind_address is not None:
+ if not utils.IsValidIP(self.op.vnc_bind_address):
+ raise errors.OpPrereqError("given VNC bind address '%s' doesn't look"
+ " like a valid IP address" %
+ self.op.vnc_bind_address)
+
+ # Xen HVM device type checks
+ if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31:
+ if self.op.hvm_nic_type not in constants.HT_HVM_VALID_NIC_TYPES:
+ raise errors.OpPrereqError("Invalid NIC type %s specified for Xen HVM"
+ " hypervisor" % self.op.hvm_nic_type)
+ if self.op.hvm_disk_type not in constants.HT_HVM_VALID_DISK_TYPES:
+ raise errors.OpPrereqError("Invalid disk type %s specified for Xen HVM"
+ " hypervisor" % self.op.hvm_disk_type)
+
if self.op.start:
self.instance_status = 'up'
else:
else:
network_port = None
+ if self.op.vnc_bind_address is None:
+ self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
+
disks = _GenerateDiskTemplate(self.cfg,
self.op.disk_template,
instance, pnode_name,
kernel_path=self.op.kernel_path,
initrd_path=self.op.initrd_path,
hvm_boot_order=self.op.hvm_boot_order,
+ hvm_acpi=self.op.hvm_acpi,
+ hvm_pae=self.op.hvm_pae,
+ hvm_cdrom_image_path=self.op.hvm_cdrom_image_path,
+ vnc_bind_address=self.op.vnc_bind_address,
+ hvm_nic_type=self.op.hvm_nic_type,
+ hvm_disk_type=self.op.hvm_disk_type,
+ auto_balance=bool(self.op.auto_balance),
)
feedback_fn("* creating instance disks...")
HTYPE = constants.HTYPE_INSTANCE
_OP_REQP = ["instance_name", "mode", "disks"]
+ def _RunAllocator(self):
+ """Compute a new secondary node using an IAllocator.
+
+ """
+ ial = IAllocator(self.cfg, self.sstore,
+ mode=constants.IALLOCATOR_MODE_RELOC,
+ name=self.op.instance_name,
+ relocate_from=[self.sec_node])
+
+ ial.Run(self.op.iallocator)
+
+ if not ial.success:
+ raise errors.OpPrereqError("Can't compute nodes using"
+ " iallocator '%s': %s" % (self.op.iallocator,
+ ial.info))
+ if len(ial.nodes) != ial.required_nodes:
+ raise errors.OpPrereqError("iallocator '%s' returned invalid number"
+ " of nodes (%s), required %s" %
+ (len(ial.nodes), ial.required_nodes))
+ self.op.remote_node = ial.nodes[0]
+ logger.ToStdout("Selected new secondary for the instance: %s" %
+ self.op.remote_node)
+
def BuildHooksEnv(self):
"""Build hooks env.
This checks that the instance is in the cluster.
"""
+ if not hasattr(self.op, "remote_node"):
+ self.op.remote_node = None
+
instance = self.cfg.GetInstanceInfo(
self.cfg.ExpandInstanceName(self.op.instance_name))
if instance is None:
self.sec_node = instance.secondary_nodes[0]
- remote_node = getattr(self.op, "remote_node", None)
+ ia_name = getattr(self.op, "iallocator", None)
+ if ia_name is not None:
+ if self.op.remote_node is not None:
+ raise errors.OpPrereqError("Give either the iallocator or the new"
+ " secondary, not both")
+ self._RunAllocator()
+
+ remote_node = self.op.remote_node
if remote_node is not None:
remote_node = self.cfg.ExpandNodeName(remote_node)
if remote_node is None:
"""
instance = self.instance
+
+ # Activate the instance disks if we're replacing them on a down instance
+ if instance.status == "down":
+ op = opcodes.OpActivateInstanceDisks(instance_name=instance.name)
+ self.proc.ChainOpCode(op)
+
if instance.disk_template == constants.DT_REMOTE_RAID1:
fn = self._ExecRR1
elif instance.disk_template == constants.DT_DRBD8:
fn = self._ExecD8Secondary
else:
raise errors.ProgrammerError("Unhandled disk replacement case")
- return fn(feedback_fn)
+
+ ret = fn(feedback_fn)
+
+ # Deactivate the instance disks if we're replacing them on a down instance
+ if instance.status == "down":
+ op = opcodes.OpDeactivateInstanceDisks(instance_name=instance.name)
+ self.proc.ChainOpCode(op)
+
+ return ret
+
+
+class LUGrowDisk(LogicalUnit):
+ """Grow a disk of an instance.
+
+ """
+ HPATH = "disk-grow"
+ HTYPE = constants.HTYPE_INSTANCE
+ _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
+
+ def BuildHooksEnv(self):
+ """Build hooks env.
+
+ This runs on the master, the primary and all the secondaries.
+
+ """
+ env = {
+ "DISK": self.op.disk,
+ "AMOUNT": self.op.amount,
+ }
+ env.update(_BuildInstanceHookEnvByObject(self.instance))
+ nl = [
+ self.sstore.GetMasterNode(),
+ self.instance.primary_node,
+ ]
+ return env, nl, nl
+
+ def CheckPrereq(self):
+ """Check prerequisites.
+
+ This checks that the instance is in the cluster.
+
+ """
+ instance = self.cfg.GetInstanceInfo(
+ self.cfg.ExpandInstanceName(self.op.instance_name))
+ if instance is None:
+ raise errors.OpPrereqError("Instance '%s' not known" %
+ self.op.instance_name)
+
+ if self.op.amount <= 0:
+ raise errors.OpPrereqError("Invalid grow-by amount: %s" % self.op.amount)
+
+ self.instance = instance
+ self.op.instance_name = instance.name
+
+ if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
+ raise errors.OpPrereqError("Instance's disk layout does not support"
+ " growing.")
+
+ self.disk = instance.FindDisk(self.op.disk)
+ if self.disk is None:
+ raise errors.OpPrereqError("Disk '%s' not found for instance '%s'" %
+ (self.op.disk, instance.name))
+
+ nodenames = [instance.primary_node] + list(instance.secondary_nodes)
+ nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName())
+ for node in nodenames:
+ info = nodeinfo.get(node, None)
+ if not info:
+ raise errors.OpPrereqError("Cannot get current information"
+ " from node '%s'" % node)
+ 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 self.op.amount > info['vg_free']:
+ raise errors.OpPrereqError("Not enough disk space on target node %s:"
+ " %d MiB available, %d MiB required" %
+ (node, info['vg_free'], self.op.amount))
+ is_primary = (node == instance.primary_node)
+ if not _CheckDiskConsistency(self.cfg, self.disk, node, is_primary):
+ raise errors.OpPrereqError("Disk %s is degraded or not fully"
+ " synchronized on node %s,"
+ " aborting grow." % (self.op.disk, node))
+
+ def Exec(self, feedback_fn):
+ """Execute disk grow.
+
+ """
+ instance = self.instance
+ disk = self.disk
+ for node in (instance.secondary_nodes + (instance.primary_node,)):
+ self.cfg.SetDiskID(disk, node)
+ result = rpc.call_blockdev_grow(node, disk, self.op.amount)
+ if not result or not isinstance(result, tuple) or len(result) != 2:
+ raise errors.OpExecError("grow request failed to node %s" % node)
+ elif not result[0]:
+ raise errors.OpExecError("grow request failed to node %s: %s" %
+ (node, result[1]))
+ disk.RecordGrow(self.op.amount)
+ self.cfg.Update(instance)
+ if self.op.wait_for_sync:
+ disk_abort = not _WaitForSync(self.cfg, instance, self.proc)
+ if disk_abort:
+ logger.Error("Warning: disk sync-ing has not returned a good status.\n"
+ " Please check the instance.")
class LUQueryInstanceData(NoHooksLU):
"""Query runtime instance data.
"""
- _OP_REQP = ["instances"]
+ _OP_REQP = ["instances", "static"]
def CheckPrereq(self):
"""Check prerequisites.
"""Compute block device status.
"""
- self.cfg.SetDiskID(dev, instance.primary_node)
- dev_pstatus = rpc.call_blockdev_find(instance.primary_node, dev)
+ static = self.op.static
+ if not static:
+ self.cfg.SetDiskID(dev, instance.primary_node)
+ dev_pstatus = rpc.call_blockdev_find(instance.primary_node, dev)
+ else:
+ dev_pstatus = None
+
if dev.dev_type in constants.LDS_DRBD:
# we change the snode then (otherwise we use the one passed in)
if dev.logical_id[0] == instance.primary_node:
else:
snode = dev.logical_id[0]
- if snode:
+ if snode and not static:
self.cfg.SetDiskID(dev, snode)
dev_sstatus = rpc.call_blockdev_find(snode, dev)
else:
"""Gather and return data"""
result = {}
for instance in self.wanted_instances:
- remote_info = rpc.call_instance_info(instance.primary_node,
- instance.name)
- if remote_info and "state" in remote_info:
- remote_state = "up"
+ if not self.op.static:
+ remote_info = rpc.call_instance_info(instance.primary_node,
+ instance.name)
+ if remote_info and "state" in remote_info:
+ remote_state = "up"
+ else:
+ remote_state = "down"
else:
- remote_state = "down"
+ remote_state = None
if instance.status == "down":
config_state = "down"
else:
"memory": instance.memory,
"nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
"disks": disks,
- "network_port": instance.network_port,
"vcpus": instance.vcpus,
- "kernel_path": instance.kernel_path,
- "initrd_path": instance.initrd_path,
- "hvm_boot_order": instance.hvm_boot_order,
+ "auto_balance": instance.auto_balance,
}
+ htkind = self.sstore.GetHypervisorType()
+ if htkind == constants.HT_XEN_PVM30:
+ idict["kernel_path"] = instance.kernel_path
+ idict["initrd_path"] = instance.initrd_path
+
+ if htkind == constants.HT_XEN_HVM31:
+ idict["hvm_boot_order"] = instance.hvm_boot_order
+ idict["hvm_acpi"] = instance.hvm_acpi
+ idict["hvm_pae"] = instance.hvm_pae
+ idict["hvm_cdrom_image_path"] = instance.hvm_cdrom_image_path
+ idict["hvm_nic_type"] = instance.hvm_nic_type
+ idict["hvm_disk_type"] = instance.hvm_disk_type
+
+ if htkind in constants.HTS_REQ_PORT:
+ if instance.network_port is None:
+ vnc_console_port = None
+ elif instance.vnc_bind_address == constants.BIND_ADDRESS_GLOBAL:
+ vnc_console_port = "%s:%s" % (instance.primary_node,
+ instance.network_port)
+ elif instance.vnc_bind_address == constants.LOCALHOST_IP_ADDRESS:
+ vnc_console_port = "%s:%s on node %s" % (instance.vnc_bind_address,
+ instance.network_port,
+ instance.primary_node)
+ else:
+ vnc_console_port = "%s:%s" % (instance.vnc_bind_address,
+ instance.network_port)
+ idict["vnc_console_port"] = vnc_console_port
+ idict["vnc_bind_address"] = instance.vnc_bind_address
+ idict["network_port"] = instance.network_port
+
result[instance.name] = idict
return result
self.kernel_path = getattr(self.op, "kernel_path", None)
self.initrd_path = getattr(self.op, "initrd_path", None)
self.hvm_boot_order = getattr(self.op, "hvm_boot_order", None)
- all_parms = [self.mem, self.vcpus, self.ip, self.bridge, self.mac,
- self.kernel_path, self.initrd_path, self.hvm_boot_order]
+ self.hvm_acpi = getattr(self.op, "hvm_acpi", None)
+ self.hvm_pae = getattr(self.op, "hvm_pae", None)
+ self.hvm_nic_type = getattr(self.op, "hvm_nic_type", None)
+ self.hvm_disk_type = getattr(self.op, "hvm_disk_type", None)
+ self.hvm_cdrom_image_path = getattr(self.op, "hvm_cdrom_image_path", None)
+ self.vnc_bind_address = getattr(self.op, "vnc_bind_address", None)
+ self.force = getattr(self.op, "force", None)
+ self.auto_balance = getattr(self.op, "auto_balance", None)
+ all_parms = [
+ self.mem, self.vcpus, self.ip, self.bridge, self.mac,
+ self.kernel_path, self.initrd_path, self.hvm_boot_order,
+ self.hvm_acpi, self.hvm_pae, self.hvm_cdrom_image_path,
+ self.vnc_bind_address, self.hvm_nic_type, self.hvm_disk_type,
+ self.auto_balance,
+ ]
if all_parms.count(None) == len(all_parms):
raise errors.OpPrereqError("No changes submitted")
if self.mem is not None:
" must be one or more of [acdn]"
" or 'default'")
+ # hvm_cdrom_image_path verification
+ if self.op.hvm_cdrom_image_path is not None:
+ if not (os.path.isabs(self.op.hvm_cdrom_image_path) or
+ self.op.hvm_cdrom_image_path.lower() == "none"):
+ raise errors.OpPrereqError("The path to the HVM CDROM image must"
+ " be an absolute path or None, not %s" %
+ self.op.hvm_cdrom_image_path)
+ if not (os.path.isfile(self.op.hvm_cdrom_image_path) or
+ self.op.hvm_cdrom_image_path.lower() == "none"):
+ raise errors.OpPrereqError("The HVM CDROM image must either be a"
+ " regular file or a symlink pointing to"
+ " an existing regular file, not %s" %
+ self.op.hvm_cdrom_image_path)
+
+ # vnc_bind_address verification
+ if self.op.vnc_bind_address is not None:
+ if not utils.IsValidIP(self.op.vnc_bind_address):
+ raise errors.OpPrereqError("given VNC bind address '%s' doesn't look"
+ " like a valid IP address" %
+ self.op.vnc_bind_address)
+
+ # Xen HVM device type checks
+ if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31:
+ if self.op.hvm_nic_type is not None:
+ if self.op.hvm_nic_type not in constants.HT_HVM_VALID_NIC_TYPES:
+ raise errors.OpPrereqError("Invalid NIC type %s specified for Xen"
+ " HVM hypervisor" % self.op.hvm_nic_type)
+ if self.op.hvm_disk_type is not None:
+ if self.op.hvm_disk_type not in constants.HT_HVM_VALID_DISK_TYPES:
+ raise errors.OpPrereqError("Invalid disk type %s specified for Xen"
+ " HVM hypervisor" % self.op.hvm_disk_type)
+
+ # auto balance setting
+ if self.auto_balance is not None:
+ # convert the value to a proper bool value, if it's not
+ self.auto_balance = bool(self.auto_balance)
+
instance = self.cfg.GetInstanceInfo(
self.cfg.ExpandInstanceName(self.op.instance_name))
if instance is None:
self.op.instance_name)
self.op.instance_name = instance.name
self.instance = instance
+ self.warn = []
+ if self.mem is not None and not self.force:
+ pnode = self.instance.primary_node
+ nodelist = [pnode]
+ if instance.auto_balance:
+ nodelist.extend(instance.secondary_nodes)
+ instance_info = rpc.call_instance_info(pnode, instance.name)
+ nodeinfo = rpc.call_node_info(nodelist, self.cfg.GetVGName())
+
+ if pnode not in nodeinfo or not isinstance(nodeinfo[pnode], dict):
+ # Assume the primary node is unreachable and go ahead
+ self.warn.append("Can't get info from primary node %s" % pnode)
+ else:
+ if instance_info:
+ current_mem = instance_info['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 = self.mem - current_mem - nodeinfo[pnode]['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 instance.auto_balance:
+ for node in instance.secondary_nodes:
+ if node not in nodeinfo or not isinstance(nodeinfo[node], dict):
+ self.warn.append("Can't get info from secondary node %s" % node)
+ elif self.mem > nodeinfo[node]['memory_free']:
+ self.warn.append("Not enough memory to failover instance to"
+ " secondary node %s" % node)
return
def Exec(self, feedback_fn):
All parameters take effect only at the next restart of the instance.
"""
+ # Process here the warnings from CheckPrereq, as we don't have a
+ # feedback_fn there.
+ for warn in self.warn:
+ feedback_fn("WARNING: %s" % warn)
+
result = []
instance = self.instance
if self.mem:
else:
instance.hvm_boot_order = self.hvm_boot_order
result.append(("hvm_boot_order", self.hvm_boot_order))
+ if self.hvm_acpi is not None:
+ instance.hvm_acpi = self.hvm_acpi
+ result.append(("hvm_acpi", self.hvm_acpi))
+ if self.hvm_pae is not None:
+ instance.hvm_pae = self.hvm_pae
+ result.append(("hvm_pae", self.hvm_pae))
+ if self.hvm_nic_type is not None:
+ instance.hvm_nic_type = self.hvm_nic_type
+ result.append(("hvm_nic_type", self.hvm_nic_type))
+ if self.hvm_disk_type is not None:
+ instance.hvm_disk_type = self.hvm_disk_type
+ result.append(("hvm_disk_type", self.hvm_disk_type))
+ if self.hvm_cdrom_image_path:
+ if self.hvm_cdrom_image_path == constants.VALUE_NONE:
+ instance.hvm_cdrom_image_path = None
+ else:
+ instance.hvm_cdrom_image_path = self.hvm_cdrom_image_path
+ result.append(("hvm_cdrom_image_path", self.hvm_cdrom_image_path))
+ if self.vnc_bind_address:
+ instance.vnc_bind_address = self.vnc_bind_address
+ result.append(("vnc_bind_address", self.vnc_bind_address))
+ if self.auto_balance is not None:
+ instance.auto_balance = self.auto_balance
+ result.append(("auto_balance", self.auto_balance))
self.cfg.AddInstance(instance)
if self.op.shutdown:
# shutdown the instance, but not the disks
if not rpc.call_instance_shutdown(src_node, instance):
- raise errors.OpExecError("Could not shutdown instance %s on node %s" %
- (instance.name, source_node))
+ raise errors.OpExecError("Could not shutdown instance %s on node %s" %
+ (instance.name, src_node))
vgname = self.cfg.GetVGName()
"""
TagsLU.CheckPrereq(self)
for tag in self.op.tags:
- objects.TaggableObject.ValidateTag(tag)
+ objects.TaggableObject.ValidateTag(tag, removal=True)
del_tags = frozenset(self.op.tags)
cur_tags = self.target.GetTags()
if not del_tags <= cur_tags:
easy usage
"""
- _KEYS = [
- "mode", "name",
+ _ALLO_KEYS = [
"mem_size", "disks", "disk_template",
"os", "tags", "nics", "vcpus",
]
+ _RELO_KEYS = [
+ "relocate_from",
+ ]
- def __init__(self, cfg, sstore, **kwargs):
+ def __init__(self, cfg, sstore, mode, name, **kwargs):
self.cfg = cfg
self.sstore = sstore
# 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
- self.mode = self.name = None
+ self.mode = mode
+ self.name = name
self.mem_size = self.disks = self.disk_template = None
self.os = self.tags = self.nics = self.vcpus = None
+ self.relocate_from = None
# computed fields
self.required_nodes = None
# init result fields
self.success = self.info = self.nodes = None
+ if self.mode == constants.IALLOCATOR_MODE_ALLOC:
+ keyset = self._ALLO_KEYS
+ elif self.mode == constants.IALLOCATOR_MODE_RELOC:
+ keyset = self._RELO_KEYS
+ else:
+ raise errors.ProgrammerError("Unknown mode '%s' passed to the"
+ " IAllocator" % self.mode)
for key in kwargs:
- if key not in self._KEYS:
+ if key not in keyset:
raise errors.ProgrammerError("Invalid input parameter '%s' to"
" IAllocator" % key)
setattr(self, key, kwargs[key])
- for key in self._KEYS:
+ for key in keyset:
if key not in kwargs:
raise errors.ProgrammerError("Missing input parameter '%s' to"
" IAllocator" % key)
"version": 1,
"cluster_name": self.sstore.GetClusterName(),
"cluster_tags": list(cfg.GetClusterInfo().GetTags()),
+ "hypervisor_type": self.sstore.GetHypervisorType(),
# we don't have job IDs
}
+ i_list = [cfg.GetInstanceInfo(iname) for iname in cfg.GetInstanceList()]
+
# node data
node_results = {}
node_list = cfg.GetNodeList()
if nname not in node_data or not isinstance(node_data[nname], dict):
raise errors.OpExecError("Can't get data for node %s" % nname)
remote_info = node_data[nname]
- for attr in ['memory_total', 'memory_free',
- 'vg_size', 'vg_free']:
+ 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:
- int(remote_info[attr])
+ remote_info[attr] = int(remote_info[attr])
except ValueError, err:
raise errors.OpExecError("Node '%s' returned invalid value for '%s':"
" %s" % (nname, attr, str(err)))
+ # compute memory used by primary instances
+ i_p_mem = i_p_up_mem = 0
+ for iinfo in i_list:
+ if iinfo.primary_node == nname:
+ i_p_mem += iinfo.memory
+ if iinfo.status == "up":
+ i_p_up_mem += iinfo.memory
+
+ # compute memory used by instances
pnr = {
"tags": list(ninfo.GetTags()),
- "total_memory": utils.TryConvert(int, remote_info['memory_total']),
- "free_memory": utils.TryConvert(int, remote_info['memory_free']),
- "total_disk": utils.TryConvert(int, remote_info['vg_size']),
- "free_disk": utils.TryConvert(int, remote_info['vg_free']),
+ "total_memory": remote_info['memory_total'],
+ "reserved_memory": remote_info['memory_dom0'],
+ "free_memory": remote_info['memory_free'],
+ "i_pri_memory": i_p_mem,
+ "i_pri_up_memory": i_p_up_mem,
+ "total_disk": remote_info['vg_size'],
+ "free_disk": remote_info['vg_free'],
"primary_ip": ninfo.primary_ip,
"secondary_ip": ninfo.secondary_ip,
+ "total_cpus": remote_info['cpu_total'],
}
node_results[nname] = pnr
data["nodes"] = node_results
# instance data
instance_data = {}
- i_list = cfg.GetInstanceList()
- for iname in i_list:
- iinfo = cfg.GetInstanceInfo(iname)
+ for iinfo in i_list:
nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
for n in iinfo.nics]
pir = {
"disks": [{"size": dsk.size, "mode": "w"} for dsk in iinfo.disks],
"disk_template": iinfo.disk_template,
}
- instance_data[iname] = pir
+ instance_data[iinfo.name] = pir
data["instances"] = instance_data
"name": self.name,
"disk_space_total": disk_space,
"required_nodes": self.required_nodes,
- "nodes": list(instance.secondary_nodes),
+ "relocate_from": self.relocate_from,
}
self.in_data["request"] = request
self.in_text = serializer.Dump(self.in_data)
- def Run(self, name, validate=True):
+ def Run(self, name, validate=True, call_fn=rpc.call_iallocator_runner):
"""Run an instance allocator and return the results.
"""
data = self.in_text
- alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
- os.path.isfile)
- if alloc_script is None:
- raise errors.OpExecError("Can't find allocator '%s'" % name)
+ result = call_fn(self.sstore.GetMasterNode(), name, self.in_text)
- fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
- try:
- os.write(fd, data)
- os.close(fd)
- result = utils.RunCmd([alloc_script, fin_name])
- if result.failed:
+ if not isinstance(result, tuple) or len(result) != 4:
+ raise errors.OpExecError("Invalid result from master iallocator runner")
+
+ rcode, stdout, stderr, fail = result
+
+ 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" %
- (result.fail_reason, result.output))
- finally:
- os.unlink(fin_name)
- self.out_text = result.stdout
+ (fail, stdout+stderr))
+ self.out_text = stdout
if validate:
self._ValidateResult()
raise errors.OpPrereqError("Instance '%s' not found for relocation" %
self.op.name)
self.op.name = fname
+ self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
else:
raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
self.op.mode)
"""Run the allocator test.
"""
- ial = IAllocator(self.cfg, self.sstore,
- mode=self.op.mode,
- name=self.op.name,
- mem_size=self.op.mem_size,
- disks=self.op.disks,
- disk_template=self.op.disk_template,
- os=self.op.os,
- tags=self.op.tags,
- nics=self.op.nics,
- vcpus=self.op.vcpus,
- )
+ if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
+ ial = IAllocator(self.cfg, self.sstore,
+ mode=self.op.mode,
+ name=self.op.name,
+ mem_size=self.op.mem_size,
+ disks=self.op.disks,
+ disk_template=self.op.disk_template,
+ os=self.op.os,
+ tags=self.op.tags,
+ nics=self.op.nics,
+ vcpus=self.op.vcpus,
+ )
+ else:
+ ial = IAllocator(self.cfg, self.sstore,
+ mode=self.op.mode,
+ name=self.op.name,
+ relocate_from=list(self.relocate_from),
+ )
if self.op.direction == constants.IALLOCATOR_DIR_IN:
result = ial.in_text