From d5cd389c61faca21f200400d2e65b2148da79ddd Mon Sep 17 00:00:00 2001
From: Iustin Pop <iustin@google.com>
Date: Mon, 15 Feb 2010 14:34:07 +0100
Subject: [PATCH] Release all node locks during disk replace
This patch extends commit 7ea7bcf by releasing all node locks in disk
replace for the early release mode. The rationale behind this is:
- LUCreateInstance already releases all node locks while waiting for
disk synchronization, and does an instance startup later
- WaitForSync only runs (for disk template 'drbd') 'lvs' and read
/proc/drbd on the primary node, which should be (modulo bugs in LVM)
safe for parallel run
In any case, the worst I could foresee is a node having N lvs commands
run in parallel on it, while being a primary for disk storage. Based on
create instance doing this safely, and the fact that burnin with more
than two instances per node is safe, I think this can be applied.
Signed-off-by: Iustin Pop <iustin@google.com>
Reviewed-by: Michael Hanselmann <hansmi@google.com>
---
lib/cmdlib.py | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/lib/cmdlib.py b/lib/cmdlib.py
index 8d4adffc5..8e2cf3a7f 100644
--- a/lib/cmdlib.py
+++ b/lib/cmdlib.py
@@ -7012,10 +7012,9 @@ class TLReplaceDisks(Tasklet):
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)
+ # WARNING: we release both node locks here, do not do other RPCs
+ # than WaitForSync to the primary node
+ self._ReleaseNodeLock([self.target_node, self.other_node])
# Wait for sync
# This can fail as the old devices are degraded and _WaitForSync
@@ -7170,7 +7169,11 @@ class TLReplaceDisks(Tasklet):
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])
+ # WARNING: we release all node locks here, do not do other RPCs
+ # than WaitForSync to the primary node
+ self._ReleaseNodeLock([self.instance.primary_node,
+ self.target_node,
+ self.new_node])
# Wait for sync
# This can fail as the old devices are degraded and _WaitForSync
--
GitLab