test = msg and not res.offline
self._ErrorIf(test, self.ENODEHOOKS, node_name,
"Communication failure in hooks execution: %s", msg)
- if test:
+ if res.offline or msg:
+ # No need to investigate payload if node is offline or gave an error.
# override manually lu_result here as _ErrorIf only
# overrides self.bad
lu_result = 1
"NEW_NAME": self.op.name,
}
mn = self.cfg.GetMasterNode()
- return env, [mn], [mn]
+ all_nodes = self.cfg.GetNodeList()
+ return env, [mn], all_nodes
def CheckPrereq(self):
"""Verify that the passed name is a valid one.
"NODE_NAME": self.op.node_name,
}
all_nodes = self.cfg.GetNodeList()
- if self.op.node_name in all_nodes:
+ try:
all_nodes.remove(self.op.node_name)
+ except ValueError:
+ logging.warning("Node %s which is about to be removed not found"
+ " in the all nodes list", self.op.node_name)
return env, all_nodes, all_nodes
def CheckPrereq(self):
# If we're being deofflined/drained, we'll MC ourself if needed
if (deoffline_or_drain and not offline_or_drain and not
- self.op.master_candidate == True):
+ self.op.master_candidate == True and not node.master_candidate):
self.op.master_candidate = _DecideSelfPromotion(self)
if self.op.master_candidate:
self.LogInfo("Autopromoting node to master candidate")
_StartInstanceDisks(self, inst, None)
try:
feedback_fn("Running the instance OS create scripts...")
- result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
+ # FIXME: pass debug option from opcode to backend
+ result = self.rpc.call_instance_os_add(inst.primary_node, inst, True, 0)
result.Raise("Could not install OS for instance %s on node %s" %
(inst.name, inst.primary_node))
finally:
_StartInstanceDisks(self, inst, None)
try:
result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
- old_name)
+ old_name, 0)
msg = result.fail_msg
if msg:
msg = ("Could not run OS rename script for instance %s on node %s"
# MAC address verification
mac = nic.get("mac", constants.VALUE_AUTO)
if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
- if not utils.IsValidMac(mac.lower()):
- raise errors.OpPrereqError("Invalid MAC address specified: %s" %
- mac, errors.ECODE_INVAL)
- else:
- try:
- self.cfg.ReserveMAC(mac, self.proc.GetECId())
- except errors.ReservationError:
- raise errors.OpPrereqError("MAC address %s already in use"
- " in cluster" % mac,
- errors.ECODE_NOTUNIQUE)
+ mac = utils.NormalizeAndValidateMac(mac)
+
+ try:
+ self.cfg.ReserveMAC(mac, self.proc.GetECId())
+ except errors.ReservationError:
+ raise errors.OpPrereqError("MAC address %s already in use"
+ " in cluster" % mac,
+ errors.ECODE_NOTUNIQUE)
# bridge verification
bridge = nic.get("bridge", None)
raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
try:
size = int(size)
- except ValueError:
+ except (TypeError, ValueError):
raise errors.OpPrereqError("Invalid disk size '%s'" % size,
errors.ECODE_INVAL)
self.disks.append({"size": size, "mode": mode})
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, False)
+ # FIXME: pass debug option from opcode to backend
+ result = self.rpc.call_instance_os_add(pnode_name, iobj, False, 0)
result.Raise("Could not add os for instance %s"
" on node %s" % (instance, pnode_name))
src_node = self.op.src_node
src_images = self.src_images
cluster_name = self.cfg.GetClusterName()
+ # FIXME: pass debug option from opcode to backend
import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
src_node, src_images,
- cluster_name)
+ cluster_name, 0)
msg = import_result.fail_msg
if msg:
self.LogWarning("Error while importing the disk images for instance"
self.op.remote_node = None
if not hasattr(self.op, "iallocator"):
self.op.iallocator = None
+ if not hasattr(self.op, "early_release"):
+ self.op.early_release = False
TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
self.op.iallocator)
self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
self.op.iallocator, self.op.remote_node,
- self.op.disks)
+ self.op.disks, False, self.op.early_release)
self.tasklets = [self.replacer]
self.op.remote_node = None
if not hasattr(self.op, "iallocator"):
self.op.iallocator = None
+ if not hasattr(self.op, "early_release"):
+ self.op.early_release = False
TLReplaceDisks.CheckArguments(constants.REPLACE_DISK_CHG,
self.op.remote_node,
names.append(inst.name)
replacer = TLReplaceDisks(self, inst.name, constants.REPLACE_DISK_CHG,
- self.op.iallocator, self.op.remote_node, [])
+ self.op.iallocator, self.op.remote_node, [],
+ True, self.op.early_release)
tasklets.append(replacer)
self.tasklets = tasklets
"""
def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
- disks):
+ disks, delay_iallocator, early_release):
"""Initializes this class.
"""
self.iallocator_name = iallocator_name
self.remote_node = remote_node
self.disks = disks
+ self.delay_iallocator = delay_iallocator
+ self.early_release = early_release
# Runtime data
self.instance = None
len(instance.secondary_nodes),
errors.ECODE_FAULT)
+ if not self.delay_iallocator:
+ self._CheckPrereq2()
+
+ def _CheckPrereq2(self):
+ """Check prerequisites, second part.
+
+ This function should always be part of CheckPrereq. It was separated and is
+ now called from Exec because during node evacuation iallocator was only
+ called with an unmodified cluster model, not taking planned changes into
+ account.
+
+ """
+ instance = self.instance
secondary_node = instance.secondary_nodes[0]
if self.iallocator_name is None:
This dispatches the disk replacement to the appropriate handler.
"""
+ if self.delay_iallocator:
+ self._CheckPrereq2()
+
if not self.disks:
feedback_fn("No disks need replacement")
return
self.lu.LogWarning("Can't remove old LV: %s" % msg,
hint="remove unused LVs manually")
+ def _ReleaseNodeLock(self, node_name):
+ """Releases the lock for a given node."""
+ self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
+
def _ExecDrbd8DiskOnly(self, feedback_fn):
"""Replace a disk on the primary or secondary for DRBD 8.
self.cfg.Update(self.instance, feedback_fn)
+ cstep = 5
+ if self.early_release:
+ self.lu.LogStep(cstep, steps_total, "Removing old storage")
+ cstep += 1
+ self._RemoveOldStorage(self.target_node, iv_names)
+ # only release the lock if we're doing secondary replace, since
+ # we use the primary node later
+ if self.target_node != self.instance.primary_node:
+ self._ReleaseNodeLock(self.target_node)
+
# Wait for sync
# This can fail as the old devices are degraded and _WaitForSync
# does a combined result over all disks, so we don't check its return value
- self.lu.LogStep(5, steps_total, "Sync devices")
+ self.lu.LogStep(cstep, steps_total, "Sync devices")
+ cstep += 1
_WaitForSync(self.lu, self.instance)
# Check all devices manually
self._CheckDevices(self.instance.primary_node, iv_names)
# Step: remove old storage
- self.lu.LogStep(6, steps_total, "Removing old storage")
- self._RemoveOldStorage(self.target_node, iv_names)
+ if not self.early_release:
+ self.lu.LogStep(cstep, steps_total, "Removing old storage")
+ cstep += 1
+ self._RemoveOldStorage(self.target_node, iv_names)
def _ExecDrbd8Secondary(self, feedback_fn):
"""Replace the secondary node for DRBD 8.
to_node, msg,
hint=("please do a gnt-instance info to see the"
" status of disks"))
+ cstep = 5
+ if self.early_release:
+ self.lu.LogStep(cstep, steps_total, "Removing old storage")
+ cstep += 1
+ self._RemoveOldStorage(self.target_node, iv_names)
+ self._ReleaseNodeLock([self.target_node, self.new_node])
# Wait for sync
# This can fail as the old devices are degraded and _WaitForSync
# does a combined result over all disks, so we don't check its return value
- self.lu.LogStep(5, steps_total, "Sync devices")
+ self.lu.LogStep(cstep, steps_total, "Sync devices")
+ cstep += 1
_WaitForSync(self.lu, self.instance)
# Check all devices manually
self._CheckDevices(self.instance.primary_node, iv_names)
# Step: remove old storage
- self.lu.LogStep(6, steps_total, "Removing old storage")
- self._RemoveOldStorage(self.target_node, iv_names)
+ if not self.early_release:
+ self.lu.LogStep(cstep, steps_total, "Removing old storage")
+ self._RemoveOldStorage(self.target_node, iv_names)
class LURepairNodeStorage(NoHooksLU):
errors.ECODE_INVAL)
try:
size = int(size)
- except ValueError, err:
+ except (TypeError, ValueError), err:
raise errors.OpPrereqError("Invalid disk size parameter: %s" %
str(err), errors.ECODE_INVAL)
disk_dict['size'] = size
if 'mac' in nic_dict:
nic_mac = nic_dict['mac']
if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
- if not utils.IsValidMac(nic_mac):
- raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac,
- errors.ECODE_INVAL)
+ nic_mac = utils.NormalizeAndValidateMac(nic_mac)
+
if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
raise errors.OpPrereqError("'auto' is not a valid MAC address when"
" modifying an existing nic",
feedback_fn("Exporting snapshot %s from %s to %s" %
(idx, src_node, dst_node.name))
if dev:
+ # FIXME: pass debug from opcode to backend
result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
- instance, cluster_name, idx)
+ instance, cluster_name,
+ idx, 0)
msg = result.fail_msg
if msg:
self.LogWarning("Could not export disk/%s from node %s to"