from ganeti import query
from ganeti import qlang
from ganeti import opcodes
+from ganeti import ht
import ganeti.masterd.instance # pylint: disable-msg=W0611
iallocator = getattr(lu.op, iallocator_slot, None)
if node is not None and iallocator is not None:
- raise errors.OpPrereqError("Do not specify both, iallocator and node.",
+ raise errors.OpPrereqError("Do not specify both, iallocator and node",
errors.ECODE_INVAL)
elif node is None and iallocator is None:
default_iallocator = lu.cfg.GetDefaultIAllocator()
setattr(lu.op, iallocator_slot, default_iallocator)
else:
raise errors.OpPrereqError("No iallocator or node given and no"
- " cluster-wide default iallocator found."
- " Please specify either an iallocator or a"
+ " cluster-wide default iallocator found;"
+ " please specify either an iallocator or a"
" node, or set a cluster-wide default"
- " iallocator.")
+ " iallocator")
class LUClusterPostInit(LogicalUnit):
def _VerifyCertificate(filename):
- """Verifies a certificate for LUClusterVerify.
+ """Verifies a certificate for LUClusterVerifyConfig.
@type filename: string
@param filename: Path to PEM file
cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
utils.ReadFile(filename))
except Exception, err: # pylint: disable-msg=W0703
- return (LUClusterVerify.ETYPE_ERROR,
+ return (LUClusterVerifyConfig.ETYPE_ERROR,
"Failed to load X509 certificate %s: %s" % (filename, err))
(errcode, msg) = \
if errcode is None:
return (None, fnamemsg)
elif errcode == utils.CERT_WARNING:
- return (LUClusterVerify.ETYPE_WARNING, fnamemsg)
+ return (LUClusterVerifyConfig.ETYPE_WARNING, fnamemsg)
elif errcode == utils.CERT_ERROR:
- return (LUClusterVerify.ETYPE_ERROR, fnamemsg)
+ return (LUClusterVerifyConfig.ETYPE_ERROR, fnamemsg)
raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
-class LUClusterVerify(LogicalUnit):
- """Verifies the cluster status.
+def _GetAllHypervisorParameters(cluster, instances):
+ """Compute the set of all hypervisor parameters.
+
+ @type cluster: L{objects.Cluster}
+ @param cluster: the cluster object
+ @param instances: list of L{objects.Instance}
+ @param instances: additional instances from which to obtain parameters
+ @rtype: list of (origin, hypervisor, parameters)
+ @return: a list with all parameters found, indicating the hypervisor they
+ apply to, and the origin (can be "cluster", "os X", or "instance Y")
"""
- HPATH = "cluster-verify"
- HTYPE = constants.HTYPE_CLUSTER
- REQ_BGL = False
+ hvp_data = []
+
+ for hv_name in cluster.enabled_hypervisors:
+ hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
+
+ for os_name, os_hvp in cluster.os_hvp.items():
+ for hv_name, hv_params in os_hvp.items():
+ if hv_params:
+ full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
+ hvp_data.append(("os %s" % os_name, hv_name, full_params))
+
+ # TODO: collapse identical parameter values in a single one
+ for instance in instances:
+ if instance.hvparams:
+ hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
+ cluster.FillHV(instance)))
+
+ return hvp_data
+
+class _VerifyErrors(object):
+ """Mix-in for cluster/group verify LUs.
+
+ It provides _Error and _ErrorIf, and updates the self.bad boolean. (Expects
+ self.op and self._feedback_fn to be available.)
+
+ """
TCLUSTER = "cluster"
TNODE = "node"
TINSTANCE = "instance"
ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
ECLUSTERFILECHECK = (TCLUSTER, "ECLUSTERFILECHECK")
+ ECLUSTERDANGLINGNODES = (TNODE, "ECLUSTERDANGLINGNODES")
+ ECLUSTERDANGLINGINST = (TNODE, "ECLUSTERDANGLINGINST")
EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
ETYPE_ERROR = "ERROR"
ETYPE_WARNING = "WARNING"
+ def _Error(self, ecode, item, msg, *args, **kwargs):
+ """Format an error message.
+
+ Based on the opcode's error_codes parameter, either format a
+ parseable error code, or a simpler error string.
+
+ This must be called only from Exec and functions called from Exec.
+
+ """
+ ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
+ itype, etxt = ecode
+ # first complete the msg
+ if args:
+ msg = msg % args
+ # then format the whole message
+ if self.op.error_codes: # This is a mix-in. pylint: disable-msg=E1101
+ msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
+ else:
+ if item:
+ item = " " + item
+ else:
+ item = ""
+ msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
+ # and finally report it via the feedback_fn
+ self._feedback_fn(" - %s" % msg) # Mix-in. pylint: disable-msg=E1101
+
+ def _ErrorIf(self, cond, *args, **kwargs):
+ """Log an error message if the passed condition is True.
+
+ """
+ cond = (bool(cond)
+ or self.op.debug_simulate_errors) # pylint: disable-msg=E1101
+ if cond:
+ self._Error(*args, **kwargs)
+ # do not mark the operation as failed for WARN cases only
+ if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
+ self.bad = self.bad or cond
+
+
+class LUClusterVerifyConfig(NoHooksLU, _VerifyErrors):
+ """Verifies the cluster config.
+
+ """
+ REQ_BGL = False
+
+ def _VerifyHVP(self, hvp_data):
+ """Verifies locally the syntax of the hypervisor parameters.
+
+ """
+ for item, hv_name, hv_params in hvp_data:
+ msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
+ (item, hv_name))
+ try:
+ hv_class = hypervisor.GetHypervisor(hv_name)
+ utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
+ hv_class.CheckParameterSyntax(hv_params)
+ except errors.GenericError, err:
+ self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
+
+ def ExpandNames(self):
+ self.all_group_info = self.cfg.GetAllNodeGroupsInfo()
+ self.all_node_info = self.cfg.GetAllNodesInfo()
+ self.all_inst_info = self.cfg.GetAllInstancesInfo()
+ self.needed_locks = {}
+
+ def Exec(self, feedback_fn):
+ """Verify integrity of cluster, performing various test on nodes.
+
+ """
+ self.bad = False
+ self._feedback_fn = feedback_fn
+
+ feedback_fn("* Verifying cluster config")
+
+ for msg in self.cfg.VerifyConfig():
+ self._ErrorIf(True, self.ECLUSTERCFG, None, msg)
+
+ feedback_fn("* Verifying cluster certificate files")
+
+ for cert_filename in constants.ALL_CERT_FILES:
+ (errcode, msg) = _VerifyCertificate(cert_filename)
+ self._ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
+
+ feedback_fn("* Verifying hypervisor parameters")
+
+ self._VerifyHVP(_GetAllHypervisorParameters(self.cfg.GetClusterInfo(),
+ self.all_inst_info.values()))
+
+ feedback_fn("* Verifying all nodes belong to an existing group")
+
+ # We do this verification here because, should this bogus circumstance
+ # occur, it would never be catched by VerifyGroup, which only acts on
+ # nodes/instances reachable from existing node groups.
+
+ dangling_nodes = set(node.name for node in self.all_node_info.values()
+ if node.group not in self.all_group_info)
+
+ dangling_instances = {}
+ no_node_instances = []
+
+ for inst in self.all_inst_info.values():
+ if inst.primary_node in dangling_nodes:
+ dangling_instances.setdefault(inst.primary_node, []).append(inst.name)
+ elif inst.primary_node not in self.all_node_info:
+ no_node_instances.append(inst.name)
+
+ pretty_dangling = [
+ "%s (%s)" %
+ (node.name,
+ utils.CommaJoin(dangling_instances.get(node.name,
+ ["no instances"])))
+ for node in dangling_nodes]
+
+ self._ErrorIf(bool(dangling_nodes), self.ECLUSTERDANGLINGNODES, None,
+ "the following nodes (and their instances) belong to a non"
+ " existing group: %s", utils.CommaJoin(pretty_dangling))
+
+ self._ErrorIf(bool(no_node_instances), self.ECLUSTERDANGLINGINST, None,
+ "the following instances have a non-existing primary-node:"
+ " %s", utils.CommaJoin(no_node_instances))
+
+ return (not self.bad, [g.name for g in self.all_group_info.values()])
+
+
+class LUClusterVerifyGroup(LogicalUnit, _VerifyErrors):
+ """Verifies the status of a node group.
+
+ """
+ HPATH = "cluster-verify"
+ HTYPE = constants.HTYPE_CLUSTER
+ REQ_BGL = False
+
_HOOKS_INDENT_RE = re.compile("^", re.M)
class NodeImage(object):
self.oslist = {}
def ExpandNames(self):
+ # This raises errors.OpPrereqError on its own:
+ self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
+
+ all_node_info = self.cfg.GetAllNodesInfo()
+ all_inst_info = self.cfg.GetAllInstancesInfo()
+
+ node_names = set(node.name
+ for node in all_node_info.values()
+ if node.group == self.group_uuid)
+
+ inst_names = [inst.name
+ for inst in all_inst_info.values()
+ if inst.primary_node in node_names]
+
+ # In Exec(), we warn about mirrored instances that have primary and
+ # secondary living in separate node groups. To fully verify that
+ # volumes for these instances are healthy, we will need to do an
+ # extra call to their secondaries. We ensure here those nodes will
+ # be locked.
+ for inst in inst_names:
+ if all_inst_info[inst].disk_template in constants.DTS_INT_MIRROR:
+ node_names.update(all_inst_info[inst].secondary_nodes)
+
self.needed_locks = {
- locking.LEVEL_NODE: locking.ALL_SET,
- locking.LEVEL_INSTANCE: locking.ALL_SET,
+ locking.LEVEL_NODEGROUP: [self.group_uuid],
+ locking.LEVEL_NODE: list(node_names),
+ locking.LEVEL_INSTANCE: inst_names,
}
+
self.share_locks = dict.fromkeys(locking.LEVELS, 1)
- def _Error(self, ecode, item, msg, *args, **kwargs):
- """Format an error message.
+ def CheckPrereq(self):
+ self.all_node_info = self.cfg.GetAllNodesInfo()
+ self.all_inst_info = self.cfg.GetAllInstancesInfo()
- Based on the opcode's error_codes parameter, either format a
- parseable error code, or a simpler error string.
+ group_nodes = set(node.name
+ for node in self.all_node_info.values()
+ if node.group == self.group_uuid)
- This must be called only from Exec and functions called from Exec.
+ group_instances = set(inst.name
+ for inst in self.all_inst_info.values()
+ if inst.primary_node in group_nodes)
- """
- ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
- itype, etxt = ecode
- # first complete the msg
- if args:
- msg = msg % args
- # then format the whole message
- if self.op.error_codes:
- msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
- else:
- if item:
- item = " " + item
- else:
- item = ""
- msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
- # and finally report it via the feedback_fn
- self._feedback_fn(" - %s" % msg)
+ unlocked_nodes = \
+ group_nodes.difference(self.glm.list_owned(locking.LEVEL_NODE))
- def _ErrorIf(self, cond, *args, **kwargs):
- """Log an error message if the passed condition is True.
+ unlocked_instances = \
+ group_instances.difference(self.glm.list_owned(locking.LEVEL_INSTANCE))
- """
- cond = bool(cond) or self.op.debug_simulate_errors
- if cond:
- self._Error(*args, **kwargs)
- # do not mark the operation as failed for WARN cases only
- if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
- self.bad = self.bad or cond
+ if unlocked_nodes:
+ raise errors.OpPrereqError("missing lock for nodes: %s" %
+ utils.CommaJoin(unlocked_nodes))
+
+ if unlocked_instances:
+ raise errors.OpPrereqError("missing lock for instances: %s" %
+ utils.CommaJoin(unlocked_instances))
+
+ self.my_node_names = utils.NiceSort(group_nodes)
+ self.my_inst_names = utils.NiceSort(group_instances)
+
+ self.my_node_info = dict((name, self.all_node_info[name])
+ for name in self.my_node_names)
+
+ self.my_inst_info = dict((name, self.all_inst_info[name])
+ for name in self.my_inst_names)
+
+ # We detect here the nodes that will need the extra RPC calls for verifying
+ # split LV volumes; they should be locked.
+ extra_lv_nodes = set()
+
+ for inst in self.my_inst_info.values():
+ if inst.disk_template in constants.DTS_INT_MIRROR:
+ group = self.my_node_info[inst.primary_node].group
+ for nname in inst.secondary_nodes:
+ if self.all_node_info[nname].group != group:
+ extra_lv_nodes.add(nname)
+
+ unlocked_lv_nodes = \
+ extra_lv_nodes.difference(self.glm.list_owned(locking.LEVEL_NODE))
+
+ if unlocked_lv_nodes:
+ raise errors.OpPrereqError("these nodes could be locked: %s" %
+ utils.CommaJoin(unlocked_lv_nodes))
+ self.extra_lv_nodes = list(extra_lv_nodes)
def _VerifyNode(self, ninfo, nresult):
"""Perform some basic validation on data returned from a node.
"instance not running on its primary node %s",
node_current)
- for node, n_img in node_image.items():
- if node != node_current:
- test = instance in n_img.instances
- _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
- "instance should not run on node %s", node)
-
diskdata = [(nname, success, status, idx)
for (nname, disks) in diskstatus.items()
for idx, (success, status) in enumerate(disks)]
self._ErrorIf(test, self.ENODEORPHANLV, node,
"volume %s is unknown", volume)
- def _VerifyOrphanInstances(self, instancelist, node_image):
- """Verify the list of running instances.
-
- This checks what instances are running but unknown to the cluster.
-
- """
- for node, n_img in node_image.items():
- for o_inst in n_img.instances:
- test = o_inst not in instancelist
- self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
- "instance %s on node %s should not exist", o_inst, node)
-
def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
"""Verify N+1 Memory Resilience.
return instdisk
- def _VerifyHVP(self, hvp_data):
- """Verifies locally the syntax of the hypervisor parameters.
-
- """
- for item, hv_name, hv_params in hvp_data:
- msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
- (item, hv_name))
- try:
- hv_class = hypervisor.GetHypervisor(hv_name)
- utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
- hv_class.CheckParameterSyntax(hv_params)
- except errors.GenericError, err:
- self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
-
def BuildHooksEnv(self):
"""Build hooks env.
the output be logged in the verify output and the verification to fail.
"""
- cfg = self.cfg
-
env = {
- "CLUSTER_TAGS": " ".join(cfg.GetClusterInfo().GetTags())
+ "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
}
env.update(("NODE_TAGS_%s" % node.name, " ".join(node.GetTags()))
- for node in cfg.GetAllNodesInfo().values())
+ for node in self.my_node_info.values())
return env
"""Build hooks nodes.
"""
- return ([], self.cfg.GetNodeList())
+ assert self.my_node_names, ("Node list not gathered,"
+ " has CheckPrereq been executed?")
+ return ([], self.my_node_names)
def Exec(self, feedback_fn):
- """Verify integrity of cluster, performing various test on nodes.
+ """Verify integrity of the node group, performing various test on nodes.
"""
# This method has too many local variables. pylint: disable-msg=R0914
_ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
verbose = self.op.verbose
self._feedback_fn = feedback_fn
- feedback_fn("* Verifying global settings")
- for msg in self.cfg.VerifyConfig():
- _ErrorIf(True, self.ECLUSTERCFG, None, msg)
-
- # Check the cluster certificates
- for cert_filename in constants.ALL_CERT_FILES:
- (errcode, msg) = _VerifyCertificate(cert_filename)
- _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
vg_name = self.cfg.GetVGName()
drbd_helper = self.cfg.GetDRBDHelper()
- hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
cluster = self.cfg.GetClusterInfo()
- nodeinfo_byname = self.cfg.GetAllNodesInfo()
- nodelist = utils.NiceSort(nodeinfo_byname.keys())
- nodeinfo = [nodeinfo_byname[nname] for nname in nodelist]
- instanceinfo = self.cfg.GetAllInstancesInfo()
- instancelist = utils.NiceSort(instanceinfo.keys())
groupinfo = self.cfg.GetAllNodeGroupsInfo()
+ hypervisors = cluster.enabled_hypervisors
+ node_data_list = [self.my_node_info[name] for name in self.my_node_names]
+
i_non_redundant = [] # Non redundant instances
i_non_a_balanced = [] # Non auto-balanced instances
n_offline = 0 # Count of offline nodes
master_node = self.master_node = self.cfg.GetMasterNode()
master_ip = self.cfg.GetMasterIP()
- # Compute the set of hypervisor parameters
- hvp_data = []
- for hv_name in hypervisors:
- hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
- for os_name, os_hvp in cluster.os_hvp.items():
- for hv_name, hv_params in os_hvp.items():
- if not hv_params:
- continue
- full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
- hvp_data.append(("os %s" % os_name, hv_name, full_params))
- # TODO: collapse identical parameter values in a single one
- for instance in instanceinfo.values():
- if not instance.hvparams:
- continue
- hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
- cluster.FillHV(instance)))
- # and verify them locally
- self._VerifyHVP(hvp_data)
+ feedback_fn("* Gathering data (%d nodes)" % len(self.my_node_names))
+
+ # We will make nodes contact all nodes in their group, and one node from
+ # every other group.
+ # TODO: should it be a *random* node, different every time?
+ online_nodes = [node.name for node in node_data_list if not node.offline]
+ other_group_nodes = {}
+
+ for name in sorted(self.all_node_info):
+ node = self.all_node_info[name]
+ if (node.group not in other_group_nodes
+ and node.group != self.group_uuid
+ and not node.offline):
+ other_group_nodes[node.group] = node.name
- feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
node_verify_param = {
constants.NV_FILELIST:
utils.UniqueSequence(filename
for files in filemap
for filename in files),
- constants.NV_NODELIST: [node.name for node in nodeinfo
- if not node.offline],
+ constants.NV_NODELIST: online_nodes + other_group_nodes.values(),
constants.NV_HYPERVISOR: hypervisors,
- constants.NV_HVPARAMS: hvp_data,
- constants.NV_NODENETTEST: [(node.name, node.primary_ip,
- node.secondary_ip) for node in nodeinfo
+ constants.NV_HVPARAMS:
+ _GetAllHypervisorParameters(cluster, self.all_inst_info.values()),
+ constants.NV_NODENETTEST: [(node.name, node.primary_ip, node.secondary_ip)
+ for node in node_data_list
if not node.offline],
constants.NV_INSTANCELIST: hypervisors,
constants.NV_VERSION: None,
default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
bridges.add(default_nicpp[constants.NIC_LINK])
- for instance in instanceinfo.values():
+ for instance in self.my_inst_info.values():
for nic in instance.nics:
full_nic = cluster.SimpleFillNIC(nic.nicparams)
if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
node_image = dict((node.name, self.NodeImage(offline=node.offline,
name=node.name,
vm_capable=node.vm_capable))
- for node in nodeinfo)
+ for node in node_data_list)
# Gather OOB paths
oob_paths = []
- for node in nodeinfo:
+ for node in self.all_node_info.values():
path = _SupportsOob(self.cfg, node)
if path and path not in oob_paths:
oob_paths.append(path)
if oob_paths:
node_verify_param[constants.NV_OOB_PATHS] = oob_paths
- for instance in instancelist:
- inst_config = instanceinfo[instance]
+ for instance in self.my_inst_names:
+ inst_config = self.my_inst_info[instance]
for nname in inst_config.all_nodes:
if nname not in node_image:
- # ghost node
gnode = self.NodeImage(name=nname)
- gnode.ghost = True
+ gnode.ghost = (nname not in self.all_node_info)
node_image[nname] = gnode
inst_config.MapLVsByNode(node_vol_should)
# time before and after executing the request, we can at least have a time
# window.
nvinfo_starttime = time.time()
- all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
+ all_nvinfo = self.rpc.call_node_verify(self.my_node_names,
+ node_verify_param,
self.cfg.GetClusterName())
+ if self.extra_lv_nodes and vg_name is not None:
+ extra_lv_nvinfo = \
+ self.rpc.call_node_verify(self.extra_lv_nodes,
+ {constants.NV_LVLIST: vg_name},
+ self.cfg.GetClusterName())
+ else:
+ extra_lv_nvinfo = {}
nvinfo_endtime = time.time()
all_drbd_map = self.cfg.ComputeDRBDMap()
- feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
- instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
+ feedback_fn("* Gathering disk information (%s nodes)" %
+ len(self.my_node_names))
+ instdisk = self._CollectDiskInfo(self.my_node_names, node_image,
+ self.my_inst_info)
feedback_fn("* Verifying configuration file consistency")
- self._VerifyFiles(_ErrorIf, nodeinfo, master_node, all_nvinfo, filemap)
+
+ # If not all nodes are being checked, we need to make sure the master node
+ # and a non-checked vm_capable node are in the list.
+ absent_nodes = set(self.all_node_info).difference(self.my_node_info)
+ if absent_nodes:
+ vf_nvinfo = all_nvinfo.copy()
+ vf_node_info = list(self.my_node_info.values())
+ additional_nodes = []
+ if master_node not in self.my_node_info:
+ additional_nodes.append(master_node)
+ vf_node_info.append(self.all_node_info[master_node])
+ # Add the first vm_capable node we find which is not included
+ for node in absent_nodes:
+ nodeinfo = self.all_node_info[node]
+ if nodeinfo.vm_capable and not nodeinfo.offline:
+ additional_nodes.append(node)
+ vf_node_info.append(self.all_node_info[node])
+ break
+ key = constants.NV_FILELIST
+ vf_nvinfo.update(self.rpc.call_node_verify(additional_nodes,
+ {key: node_verify_param[key]},
+ self.cfg.GetClusterName()))
+ else:
+ vf_nvinfo = all_nvinfo
+ vf_node_info = self.my_node_info.values()
+
+ self._VerifyFiles(_ErrorIf, vf_node_info, master_node, vf_nvinfo, filemap)
feedback_fn("* Verifying node status")
refos_img = None
- for node_i in nodeinfo:
+ for node_i in node_data_list:
node = node_i.name
nimg = node_image[node]
if nimg.vm_capable:
self._VerifyNodeLVM(node_i, nresult, vg_name)
- self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
+ self._VerifyNodeDrbd(node_i, nresult, self.all_inst_info, drbd_helper,
all_drbd_map)
self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
self._UpdateNodeInstances(node_i, nresult, nimg)
self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
self._UpdateNodeOS(node_i, nresult, nimg)
+
if not nimg.os_fail:
if refos_img is None:
refos_img = nimg
self._VerifyNodeOS(node_i, nimg, refos_img)
self._VerifyNodeBridges(node_i, nresult, bridges)
+ # Check whether all running instancies are primary for the node. (This
+ # can no longer be done from _VerifyInstance below, since some of the
+ # wrong instances could be from other node groups.)
+ non_primary_inst = set(nimg.instances).difference(nimg.pinst)
+
+ for inst in non_primary_inst:
+ test = inst in self.all_inst_info
+ _ErrorIf(test, self.EINSTANCEWRONGNODE, inst,
+ "instance should not run on node %s", node_i.name)
+ _ErrorIf(not test, self.ENODEORPHANINSTANCE, node_i.name,
+ "node is running unknown instance %s", inst)
+
+ for node, result in extra_lv_nvinfo.items():
+ self._UpdateNodeVolumes(self.all_node_info[node], result.payload,
+ node_image[node], vg_name)
+
feedback_fn("* Verifying instance status")
- for instance in instancelist:
+ for instance in self.my_inst_names:
if verbose:
feedback_fn("* Verifying instance %s" % instance)
- inst_config = instanceinfo[instance]
+ inst_config = self.my_inst_info[instance]
self._VerifyInstance(instance, inst_config, node_image,
instdisk[instance])
inst_nodes_offline = []
instance_groups = {}
for node in instance_nodes:
- instance_groups.setdefault(nodeinfo_byname[node].group,
+ instance_groups.setdefault(self.all_node_info[node].group,
[]).append(node)
pretty_list = [
feedback_fn("* Verifying orphan volumes")
reserved = utils.FieldSet(*cluster.reserved_lvs)
- self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
- feedback_fn("* Verifying orphan instances")
- self._VerifyOrphanInstances(instancelist, node_image)
+ # We will get spurious "unknown volume" warnings if any node of this group
+ # is secondary for an instance whose primary is in another group. To avoid
+ # them, we find these instances and add their volumes to node_vol_should.
+ for inst in self.all_inst_info.values():
+ for secondary in inst.secondary_nodes:
+ if (secondary in self.my_node_info
+ and inst.name not in self.my_inst_info):
+ inst.MapLVsByNode(node_vol_should)
+ break
+
+ self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
feedback_fn("* Verifying N+1 Memory redundancy")
- self._VerifyNPlusOneMemory(node_image, instanceinfo)
+ self._VerifyNPlusOneMemory(node_image, self.my_inst_info)
feedback_fn("* Other Notes")
if i_non_redundant:
instance = self.instance
force = self.op.force
- self.cfg.MarkInstanceUp(instance.name)
+ if not self.op.no_remember:
+ self.cfg.MarkInstanceUp(instance.name)
if self.primary_offline:
assert self.op.ignore_offline_nodes
node_current = instance.primary_node
timeout = self.op.timeout
- self.cfg.MarkInstanceDown(instance.name)
+ if not self.op.no_remember:
+ self.cfg.MarkInstanceDown(instance.name)
if self.primary_offline:
assert self.op.ignore_offline_nodes
self.be_inst = i_bedict # the new dict (without defaults)
else:
self.be_new = self.be_inst = {}
+ be_old = cluster.FillBE(instance)
# osparams processing
if self.op.osparams:
self.warn = []
- if constants.BE_MEMORY in self.op.beparams and not self.op.force:
+ if (constants.BE_MEMORY in self.op.beparams and not self.op.force and
+ be_new[constants.BE_MEMORY] > be_old[constants.BE_MEMORY]):
mem_check_list = [pnode]
if be_new[constants.BE_AUTO_BALANCE]:
# either we changed auto_balance to yes or it was from before
for node, nres in nodeinfo.items():
if node not in instance.secondary_nodes:
continue
- msg = nres.fail_msg
- if msg:
- self.warn.append("Can't get info from secondary node %s: %s" %
- (node, msg))
- elif not isinstance(nres.payload.get('memory_free', None), int):
- self.warn.append("Secondary node %s didn't return free"
- " memory information" % node)
+ nres.Raise("Can't get info from secondary node %s" % node,
+ prereq=True, ecode=errors.ECODE_STATE)
+ if not isinstance(nres.payload.get('memory_free', None), int):
+ raise errors.OpPrereqError("Secondary node %s didn't return free"
+ " memory information" % node,
+ errors.ECODE_STATE)
elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
- self.warn.append("Not enough memory to failover instance to"
- " secondary node %s" % node)
+ raise errors.OpPrereqError("This change will prevent the instance"
+ " from failover to its secondary node"
+ " %s, due to not enough memory" % node,
+ errors.ECODE_STATE)
# NIC processing
self.nic_pnew = {}
self.success = self.info = self.result = None
try:
- (fn, keyset) = self._MODE_DATA[self.mode]
+ (fn, keyset, self._result_check) = self._MODE_DATA[self.mode]
except KeyError:
raise errors.ProgrammerError("Unknown mode '%s' passed to the"
" IAllocator" % self.mode)
constants.IALLOCATOR_MODE_ALLOC:
(_AddNewInstance,
["name", "mem_size", "disks", "disk_template", "os", "tags", "nics",
- "vcpus", "hypervisor"]),
+ "vcpus", "hypervisor"], ht.TList),
constants.IALLOCATOR_MODE_RELOC:
- (_AddRelocateInstance, ["name", "relocate_from"]),
+ (_AddRelocateInstance, ["name", "relocate_from"], ht.TList),
constants.IALLOCATOR_MODE_MEVAC:
- (_AddEvacuateNodes, ["evac_nodes"]),
+ (_AddEvacuateNodes, ["evac_nodes"],
+ ht.TListOf(ht.TAnd(ht.TIsLength(2),
+ ht.TListOf(ht.TString)))),
constants.IALLOCATOR_MODE_MRELOC:
- (_AddMultiRelocate, ["instances", "reloc_mode", "target_groups"]),
+ (_AddMultiRelocate, ["instances", "reloc_mode", "target_groups"],
+ ht.TListOf(ht.TListOf(ht.TStrictDict(True, False, {
+ # pylint: disable-msg=E1101
+ # Class '...' has no 'OP_ID' member
+ "OP_ID": ht.TElemOf([opcodes.OpInstanceFailover.OP_ID,
+ opcodes.OpInstanceMigrate.OP_ID,
+ opcodes.OpInstanceReplaceDisks.OP_ID])
+ })))),
}
def Run(self, name, validate=True, call_fn=None):
" missing key '%s'" % key)
setattr(self, key, rdict[key])
- if not isinstance(rdict["result"], list):
- raise errors.OpExecError("Can't parse iallocator results: 'result' key"
- " is not a list")
-
- if self.mode == constants.IALLOCATOR_MODE_RELOC:
- assert self.relocate_from is not None
- assert self.required_nodes == 1
+ if not self._result_check(self.result):
+ raise errors.OpExecError("Iallocator returned invalid result,"
+ " expected %s, got %s" %
+ (self._result_check, self.result),
+ errors.ECODE_INVAL)
+ if self.mode in (constants.IALLOCATOR_MODE_RELOC,
+ constants.IALLOCATOR_MODE_MEVAC):
node2group = dict((name, ndata["group"])
for (name, ndata) in self.in_data["nodes"].items())
fn = compat.partial(self._NodesToGroups, node2group,
self.in_data["nodegroups"])
- request_groups = fn(self.relocate_from)
- result_groups = fn(rdict["result"])
-
- if result_groups != request_groups:
- raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
- " differ from original groups (%s)" %
- (utils.CommaJoin(result_groups),
- utils.CommaJoin(request_groups)))
+ if self.mode == constants.IALLOCATOR_MODE_RELOC:
+ assert self.relocate_from is not None
+ assert self.required_nodes == 1
+
+ request_groups = fn(self.relocate_from)
+ result_groups = fn(rdict["result"])
+
+ if result_groups != request_groups:
+ raise errors.OpExecError("Groups of nodes returned by iallocator (%s)"
+ " differ from original groups (%s)" %
+ (utils.CommaJoin(result_groups),
+ utils.CommaJoin(request_groups)))
+ elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
+ request_groups = fn(self.evac_nodes)
+ for (instance_name, secnode) in self.result:
+ result_groups = fn([secnode])
+ if result_groups != request_groups:
+ raise errors.OpExecError("Iallocator returned new secondary node"
+ " '%s' (group '%s') for instance '%s'"
+ " which is not in original group '%s'" %
+ (secnode, utils.CommaJoin(result_groups),
+ instance_name,
+ utils.CommaJoin(request_groups)))
+ else:
+ raise errors.ProgrammerError("Unhandled mode '%s'" % self.mode)
self.out_data = rdict
projects / ganeti-local / blobdiff