From 44485f49ac9d7e4169f81a28545d1aec5955d01e Mon Sep 17 00:00:00 2001
From: Guido Trotter <ultrotter@google.com>
Date: Thu, 1 Oct 2009 17:13:41 +0100
Subject: [PATCH] Fix master candidate removal
Currently during a master candidate removal, when it's possible to
promote another node, the removal operation fails because of a corrupt
config before it's even possible to do the promotion. Fixing this by
doing the promotion before, excluding the current node.
Signed-off-by: Guido Trotter <ultrotter@google.com>
Reviewed-by: Olivier Tharan <olive@google.com>
---
lib/cmdlib.py | 13 ++++++-------
lib/config.py | 9 ++++++---
2 files changed, 12 insertions(+), 10 deletions(-)
diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index e1056903c..3ce419c95 100644
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -670,17 +670,17 @@ def _BuildInstanceHookEnvByObject(lu, instance, override=None):
return _BuildInstanceHookEnv(**args)
-def _AdjustCandidatePool(lu):
+def _AdjustCandidatePool(lu, exceptions):
"""Adjust the candidate pool after node operations.
"""
- mod_list = lu.cfg.MaintainCandidatePool()
+ mod_list = lu.cfg.MaintainCandidatePool(exceptions)
if mod_list:
lu.LogInfo("Promoted nodes to master candidate role: %s",
", ".join(node.name for node in mod_list))
for name in mod_list:
lu.context.ReaddNode(name)
- mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats()
+ mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
if mc_now > mc_max:
lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
(mc_now, mc_max))
@@ -1946,7 +1946,7 @@ class LUSetClusterParams(LogicalUnit):
if self.op.candidate_pool_size is not None:
self.cluster.candidate_pool_size = self.op.candidate_pool_size
# we need to update the pool size here, otherwise the save will fail
- _AdjustCandidatePool(self)
+ _AdjustCandidatePool(self, [])
self.cfg.Update(self.cluster)
@@ -2282,6 +2282,8 @@ class LURemoveNode(LogicalUnit):
logging.info("Stopping the node daemon and removing configs from node %s",
node.name)
+ # Promote nodes to master candidate as needed
+ _AdjustCandidatePool(self, exceptions=[node.name])
self.context.RemoveNode(node.name)
# Run post hooks on the node before it's removed
@@ -2297,9 +2299,6 @@ class LURemoveNode(LogicalUnit):
self.LogWarning("Errors encountered on the remote node while leaving"
" the cluster: %s", msg)
- # Promote nodes to master candidate as needed
- _AdjustCandidatePool(self)
-
class LUQueryNodes(NoHooksLU):
"""Logical unit for querying nodes.
diff --git a/lib/config.py b/lib/config.py
index 41fcf094d..54f80702a 100644
--- a/lib/config.py
+++ b/lib/config.py
@@ -1044,14 +1044,16 @@ class ConfigWriter:
return self._UnlockedGetMasterCandidateStats(exceptions)
@locking.ssynchronized(_config_lock)
- def MaintainCandidatePool(self):
+ def MaintainCandidatePool(self, exceptions):
"""Try to grow the candidate pool to the desired size.
+ @type exceptions: list
+ @param exceptions: if passed, list of nodes that should be ignored
@rtype: list
@return: list with the adjusted nodes (L{objects.Node} instances)
"""
- mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
+ mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(exceptions)
mod_list = []
if mc_now < mc_max:
node_list = self._config_data.nodes.keys()
@@ -1060,7 +1062,8 @@ class ConfigWriter:
if mc_now >= mc_max:
break
node = self._config_data.nodes[name]
- if node.master_candidate or node.offline or node.drained:
+ if (node.master_candidate or node.offline or node.drained or
+ node.name in exceptions):
continue
mod_list.append(node)
node.master_candidate = True
--
GitLab