#
#
-# Copyright (C) 2006, 2007 Google Inc.
+# Copyright (C) 2006, 2007, 2008, 2009, 2010 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
"""
+# pylint: disable-msg=R0904
+# R0904: Too many public methods
+
import os
import random
import logging
from ganeti import rpc
from ganeti import objects
from ganeti import serializer
+from ganeti import uidpool
+from ganeti import netutils
+from ganeti import runtime
-_config_lock = locking.SharedLock()
+_config_lock = locking.SharedLock("ConfigWriter")
# job id used for resource management at config upgrade time
-_UPGRADE_CONFIG_JID="jid-cfg-upgrade"
+_UPGRADE_CONFIG_JID = "jid-cfg-upgrade"
def _ValidateConfig(data):
class ConfigWriter:
"""The interface to the cluster configuration.
+ @ivar _temporary_lvs: reservation manager for temporary LVs
+ @ivar _all_rms: a list of all temporary reservation managers
+
"""
- def __init__(self, cfg_file=None, offline=False):
+ def __init__(self, cfg_file=None, offline=False, _getents=runtime.GetEnts):
self.write_count = 0
self._lock = _config_lock
self._config_data = None
self._cfg_file = constants.CLUSTER_CONF_FILE
else:
self._cfg_file = cfg_file
+ self._getents = _getents
self._temporary_ids = TemporaryReservationManager()
self._temporary_drbds = {}
self._temporary_macs = TemporaryReservationManager()
+ self._temporary_secrets = TemporaryReservationManager()
+ self._temporary_lvs = TemporaryReservationManager()
+ self._all_rms = [self._temporary_ids, self._temporary_macs,
+ self._temporary_secrets, self._temporary_lvs]
# Note: in order to prevent errors when resolving our name in
# _DistributeConfig, we compute it here once and reuse it; it's
# better to raise an error before starting to modify the config
# file than after it was modified
- self._my_hostname = utils.HostInfo().name
+ self._my_hostname = netutils.Hostname.GetSysName()
self._last_cluster_serial = -1
self._OpenConfig()
self._temporary_macs.Reserve(mac, ec_id)
@locking.ssynchronized(_config_lock, shared=1)
- def GenerateDRBDSecret(self):
+ def ReserveLV(self, lv_name, ec_id):
+ """Reserve an VG/LV pair for an instance.
+
+ @type lv_name: string
+ @param lv_name: the logical volume name to reserve
+
+ """
+ all_lvs = self._AllLVs()
+ if lv_name in all_lvs:
+ raise errors.ReservationError("LV already in use")
+ else:
+ self._temporary_lvs.Reserve(lv_name, ec_id)
+
+ @locking.ssynchronized(_config_lock, shared=1)
+ def GenerateDRBDSecret(self, ec_id):
"""Generate a DRBD secret.
This checks the current disks for duplicates.
"""
- all_secrets = self._AllDRBDSecrets()
- retries = 64
- while retries > 0:
- secret = utils.GenerateSecret()
- if secret not in all_secrets:
- break
- retries -= 1
- else:
- raise errors.ConfigurationError("Can't generate unique DRBD secret")
- return secret
+ return self._temporary_secrets.Generate(self._AllDRBDSecrets(),
+ utils.GenerateSecret,
+ ec_id)
def _AllLVs(self):
"""Compute the list of all LVs.
if invalid_hvs:
result.append("enabled hypervisors contains invalid entries: %s" %
invalid_hvs)
+ missing_hvp = (set(data.cluster.enabled_hypervisors) -
+ set(data.cluster.hvparams.keys()))
+ if missing_hvp:
+ result.append("hypervisor parameters missing for the enabled"
+ " hypervisor(s) %s" % utils.CommaJoin(missing_hvp))
if data.cluster.master_node not in data.nodes:
result.append("cluster has invalid primary node '%s'" %
# per-instance checks
for instance_name in data.instances:
instance = data.instances[instance_name]
+ if instance.name != instance_name:
+ result.append("instance '%s' is indexed by wrong name '%s'" %
+ (instance.name, instance_name))
if instance.primary_node not in data.nodes:
result.append("instance '%s' has invalid primary node '%s'" %
(instance_name, instance.primary_node))
for pnum in keys:
pdata = ports[pnum]
if len(pdata) > 1:
- txt = ", ".join(["%s/%s" % val for val in pdata])
+ txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
# highest used tcp port check
(mc_now, mc_max))
# node checks
- for node in data.nodes.values():
+ for node_name, node in data.nodes.items():
+ if node.name != node_name:
+ result.append("Node '%s' is indexed by wrong name '%s'" %
+ (node.name, node_name))
if [node.master_candidate, node.drained, node.offline].count(True) > 1:
result.append("Node %s state is invalid: master_candidate=%s,"
" drain=%s, offline=%s" %
node.offline))
# drbd minors check
- d_map, duplicates = self._UnlockedComputeDRBDMap()
+ _, duplicates = self._UnlockedComputeDRBDMap()
for node, minor, instance_a, instance_b in duplicates:
result.append("DRBD minor %d on node %s is assigned twice to instances"
" %s and %s" % (minor, node, instance_a, instance_b))
# IP checks
- ips = { data.cluster.master_ip: ["cluster_ip"] }
- def _helper(ip, name):
- if ip in ips:
- ips[ip].append(name)
- else:
- ips[ip] = [name]
+ default_nicparams = data.cluster.nicparams[constants.PP_DEFAULT]
+ ips = {}
+
+ def _AddIpAddress(ip, name):
+ ips.setdefault(ip, []).append(name)
+
+ _AddIpAddress(data.cluster.master_ip, "cluster_ip")
for node in data.nodes.values():
- _helper(node.primary_ip, "node:%s/primary" % node.name)
+ _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
if node.secondary_ip != node.primary_ip:
- _helper(node.secondary_ip, "node:%s/secondary" % node.name)
+ _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
+
+ for instance in data.instances.values():
+ for idx, nic in enumerate(instance.nics):
+ if nic.ip is None:
+ continue
+
+ nicparams = objects.FillDict(default_nicparams, nic.nicparams)
+ nic_mode = nicparams[constants.NIC_MODE]
+ nic_link = nicparams[constants.NIC_LINK]
+
+ if nic_mode == constants.NIC_MODE_BRIDGED:
+ link = "bridge:%s" % nic_link
+ elif nic_mode == constants.NIC_MODE_ROUTED:
+ link = "route:%s" % nic_link
+ else:
+ raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
+
+ _AddIpAddress("%s/%s" % (link, nic.ip),
+ "instance:%s/nic:%d" % (instance.name, idx))
for ip, owners in ips.items():
if len(owners) > 1:
result.append("IP address %s is used by multiple owners: %s" %
- (ip, ", ".join(owners)))
+ (ip, utils.CommaJoin(owners)))
+
return result
@locking.ssynchronized(_config_lock, shared=1)
"""
return self._config_data.cluster.rsahostkeypub
+ @locking.ssynchronized(_config_lock, shared=1)
+ def GetDefaultIAllocator(self):
+ """Get the default instance allocator for this cluster.
+
+ """
+ return self._config_data.cluster.default_iallocator
+
+ @locking.ssynchronized(_config_lock, shared=1)
+ def GetPrimaryIPFamily(self):
+ """Get cluster primary ip family.
+
+ @return: primary ip family
+
+ """
+ return self._config_data.cluster.primary_ip_family
+
@locking.ssynchronized(_config_lock)
def AddInstance(self, instance, ec_id):
"""Add an instance to the config.
"""
if not item.uuid:
item.uuid = self._GenerateUniqueID(ec_id)
- elif item.uuid in self._AllIDs(temporary=True):
- raise errors.ConfigurationError("Cannot add '%s': UUID already in use" %
- (item.name, item.uuid))
+ elif item.uuid in self._AllIDs(include_temporary=True):
+ raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
+ " in use" % (item.name, item.uuid))
def _SetInstanceStatus(self, instance_name, status):
"""Set the instance's status to a given value.
# rename the file paths in logical and physical id
file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
disk.physical_id = disk.logical_id = (disk.logical_id[0],
- os.path.join(file_storage_dir,
- inst.name,
- disk.iv_name))
+ utils.PathJoin(file_storage_dir,
+ inst.name,
+ disk.iv_name))
self._config_data.instances[inst.name] = inst
self._WriteConfig()
return self._config_data.nodes[node_name]
-
@locking.ssynchronized(_config_lock, shared=1)
def GetNodeInfo(self, node_name):
"""Get the configuration of a node, as stored in the config.
"""
return self._config_data.nodes.keys()
-
@locking.ssynchronized(_config_lock, shared=1)
def GetNodeList(self):
"""Return the list of nodes which are in the configuration.
"""
return self._UnlockedGetNodeList()
- @locking.ssynchronized(_config_lock, shared=1)
- def GetOnlineNodeList(self):
+ def _UnlockedGetOnlineNodeList(self):
"""Return the list of nodes which are online.
"""
return [node.name for node in all_nodes if not node.offline]
@locking.ssynchronized(_config_lock, shared=1)
+ def GetOnlineNodeList(self):
+ """Return the list of nodes which are online.
+
+ """
+ return self._UnlockedGetOnlineNodeList()
+
+ @locking.ssynchronized(_config_lock, shared=1)
def GetAllNodesInfo(self):
"""Get the configuration of all nodes.
This is because some data elements need uniqueness across the
whole configuration, etc.
- @warning: this function will call L{_WriteConfig()}, so it needs
- to either be called with the lock held or from a safe place
- (the constructor)
+ @warning: this function will call L{_WriteConfig()}, but also
+ L{DropECReservations} so it needs to be called only from a
+ "safe" place (the constructor). If one wanted to call it with
+ the lock held, a DropECReservationUnlocked would need to be
+ created first, to avoid causing deadlock.
"""
modified = False
# only called at config init time, without the lock held
self.DropECReservations(_UPGRADE_CONFIG_JID)
-
def _DistributeConfig(self, feedback_fn):
"""Distribute the configuration to the other nodes.
config_errors = self._UnlockedVerifyConfig()
if config_errors:
errmsg = ("Configuration data is not consistent: %s" %
- (", ".join(config_errors)))
+ (utils.CommaJoin(config_errors)))
logging.critical(errmsg)
if feedback_fn:
feedback_fn(errmsg)
self._BumpSerialNo()
txt = serializer.Dump(self._config_data.ToDict())
- utils.WriteFile(destination, data=txt)
+ getents = self._getents()
+ utils.WriteFile(destination, data=txt, gid=getents.confd_gid, mode=0640)
self.write_count += 1
if self._last_cluster_serial < self._config_data.cluster.serial_no:
if not self._offline:
result = rpc.RpcRunner.call_write_ssconf_files(
- self._UnlockedGetNodeList(),
+ self._UnlockedGetOnlineNodeList(),
self._UnlockedGetSsconfValues())
for nname, nresu in result.items():
cluster = self._config_data.cluster
cluster_tags = fn(cluster.GetTags())
+
+ hypervisor_list = fn(cluster.enabled_hypervisors)
+
+ uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
+
return {
constants.SS_CLUSTER_NAME: cluster.cluster_name,
constants.SS_CLUSTER_TAGS: cluster_tags,
constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
constants.SS_OFFLINE_NODES: off_data,
constants.SS_ONLINE_NODES: on_data,
+ constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
constants.SS_INSTANCE_LIST: instance_data,
constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
+ constants.SS_HYPERVISOR_LIST: hypervisor_list,
+ constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
+ constants.SS_UID_POOL: uid_pool,
}
@locking.ssynchronized(_config_lock, shared=1)
+ def GetSsconfValues(self):
+ """Wrapper using lock around _UnlockedGetSsconf().
+
+ """
+ return self._UnlockedGetSsconfValues()
+
+ @locking.ssynchronized(_config_lock, shared=1)
def GetVGName(self):
"""Return the volume group name.
self._WriteConfig()
@locking.ssynchronized(_config_lock, shared=1)
+ def GetDRBDHelper(self):
+ """Return DRBD usermode helper.
+
+ """
+ return self._config_data.cluster.drbd_usermode_helper
+
+ @locking.ssynchronized(_config_lock)
+ def SetDRBDHelper(self, drbd_helper):
+ """Set DRBD usermode helper.
+
+ """
+ self._config_data.cluster.drbd_usermode_helper = drbd_helper
+ self._config_data.cluster.serial_no += 1
+ self._WriteConfig()
+
+ @locking.ssynchronized(_config_lock, shared=1)
def GetMACPrefix(self):
"""Return the mac prefix.
"""
return self._config_data.cluster
+ @locking.ssynchronized(_config_lock, shared=1)
+ def HasAnyDiskOfType(self, dev_type):
+ """Check if in there is at disk of the given type in the configuration.
+
+ """
+ return self._config_data.HasAnyDiskOfType(dev_type)
+
@locking.ssynchronized(_config_lock)
def Update(self, target, feedback_fn):
"""Notify function to be called after updates.
"""Drop per-execution-context reservations
"""
- self._temporary_ids.DropECReservations(ec_id)
- self._temporary_macs.DropECReservations(ec_id)
-
+ for rm in self._all_rms:
+ rm.DropECReservations(ec_id)
projects / ganeti-local / blobdiff