def _VerifyNode(self, nodeinfo, file_list, local_cksum,
node_result, feedback_fn, master_files,
- drbd_map):
+ drbd_map, vg_name):
"""Run multiple tests against a node.
Test list:
@param drbd_map: the useddrbd minors for this node, in
form of minor: (instance, must_exist) which correspond to instances
and their running status
+ @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
"""
node = nodeinfo.name
(constants.RELEASE_VERSION, node, remote_version[1]))
# checks vg existence and size > 20G
-
- vglist = node_result.get(constants.NV_VGLIST, None)
- if not vglist:
- feedback_fn(" - ERROR: unable to check volume groups on node %s." %
- (node,))
- bad = True
- else:
- vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
- constants.MIN_VG_SIZE)
- if vgstatus:
- feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
+ if vg_name is not None:
+ vglist = node_result.get(constants.NV_VGLIST, None)
+ if not vglist:
+ feedback_fn(" - ERROR: unable to check volume groups on node %s." %
+ (node,))
bad = True
+ else:
+ vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
+ constants.MIN_VG_SIZE)
+ if vgstatus:
+ feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node))
+ bad = True
# checks config file checksum
(hv_name, hv_result))
# check used drbd list
- used_minors = node_result.get(constants.NV_DRBDLIST, [])
- if not isinstance(used_minors, (tuple, list)):
- feedback_fn(" - ERROR: cannot parse drbd status file: %s" %
- str(used_minors))
- else:
- for minor, (iname, must_exist) in drbd_map.items():
- if minor not in used_minors and must_exist:
- feedback_fn(" - ERROR: drbd minor %d of instance %s is not active" %
- (minor, iname))
- bad = True
- for minor in used_minors:
- if minor not in drbd_map:
- feedback_fn(" - ERROR: unallocated drbd minor %d is in use" % minor)
- bad = True
+ if vg_name is not None:
+ used_minors = node_result.get(constants.NV_DRBDLIST, [])
+ if not isinstance(used_minors, (tuple, list)):
+ feedback_fn(" - ERROR: cannot parse drbd status file: %s" %
+ str(used_minors))
+ else:
+ for minor, (iname, must_exist) in drbd_map.items():
+ if minor not in used_minors and must_exist:
+ feedback_fn(" - ERROR: drbd minor %d of instance %s is"
+ " not active" % (minor, iname))
+ bad = True
+ for minor in used_minors:
+ if minor not in drbd_map:
+ feedback_fn(" - ERROR: unallocated drbd minor %d is in use" %
+ minor)
+ bad = True
return bad
"""
all_nodes = self.cfg.GetNodeList()
- # TODO: populate the environment with useful information for verify hooks
- env = {}
+ env = {
+ "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
+ }
+ for node in self.cfg.GetAllNodesInfo().values():
+ env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
+
return env, [], all_nodes
def Exec(self, feedback_fn):
constants.NV_NODENETTEST: [(node.name, node.primary_ip,
node.secondary_ip) for node in nodeinfo
if not node.offline],
- constants.NV_LVLIST: vg_name,
constants.NV_INSTANCELIST: hypervisors,
- constants.NV_VGLIST: None,
constants.NV_VERSION: None,
constants.NV_HVINFO: self.cfg.GetHypervisorType(),
- constants.NV_DRBDLIST: None,
}
+ if vg_name is not None:
+ node_verify_param[constants.NV_VGLIST] = None
+ node_verify_param[constants.NV_LVLIST] = vg_name
+ node_verify_param[constants.NV_DRBDLIST] = None
all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
self.cfg.GetClusterName())
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)
+ node_drbd, vg_name)
bad = bad or result
lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
- if isinstance(lvdata, basestring):
+ if vg_name is None:
+ node_volume[node] = {}
+ elif isinstance(lvdata, basestring):
feedback_fn(" - ERROR: LVM problem on node %s: %s" %
(node, utils.SafeEncode(lvdata)))
bad = True
try:
node_info[node] = {
"mfree": int(nodeinfo['memory_free']),
- "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
"pinst": [],
"sinst": [],
# dictionary holding all instances this node is secondary for,
# secondary.
"sinst-by-pnode": {},
}
- except ValueError:
- feedback_fn(" - ERROR: invalid value returned from node %s" % (node,))
+ # 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, KeyError):
+ feedback_fn(" - ERROR: invalid nodeinfo value returned"
+ " from node %s" % (node,))
bad = True
continue
if isinstance(lvs, basestring):
logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
res_nlvm[node] = lvs
+ continue
elif not isinstance(lvs, dict):
logging.warning("Connection to node %s failed or invalid data"
" returned", node)
_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:
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.
@param rlist: a map with node names as keys and OS objects as values
@rtype: dict
- @returns: a dictionary with osnames as keys and as value another map, with
+ @return: a dictionary with osnames as keys and as value another map, with
nodes as keys and list of OS objects as values, eg::
{"debian-etch": {"node1": [<object>,...],
"""
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")
((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
_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"
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):
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"
try:
_CreateSingleBlockDev(self, new_node, instance, new_drbd,
_GetInstanceInfoText(instance), False)
- except errors.BlockDeviceError:
+ except errors.GenericError:
self.cfg.ReleaseDRBDMinors(instance.name)
raise
cluster_info = cfg.GetClusterInfo()
# cluster data
data = {
- "version": 1,
+ "version": constants.IALLOCATOR_VERSION,
"cluster_name": cfg.GetClusterName(),
"cluster_tags": list(cluster_info.GetTags()),
"enabled_hypervisors": list(cluster_info.enabled_hypervisors),