Hello Kudu Jenkins,
I'd like you to reexamine a change. Please visit
http://gerrit.cloudera.org:8080/7883
to look at the new patch set (#2).
Change subject: KUDU-2118. Fully shut down TabletReplica on delete
......................................................................
KUDU-2118. Fully shut down TabletReplica on delete
After the various lifecycle changes implemented in
5bca7d8ba185d62952fb3e3163cbe88d20453da0 we were left with a case where
we fail to fully shut down TABLET_DATA_DELETED replicas before
dereferencing them. Because LeaderElection via ElectionDecisionCallback
will hold a reference to RaftConsensus while a vote is taking place, the
leader election callback running on the reactor thread may destroy the
RaftConsensus object. If that occurs without RaftConsensus already being
in a fully-shutdown state, the RaftConsensus destructor runs Shutdown()
before destoying the object, which takes a lock and triggers an
assertion via AssertWaitAllowed(). This crashes the server.
This patch fixes the issue by calling Shutdown() on TABLET_DATA_DELETED
replicas before dereferencing them in TSTabletManager.
A concurrency test is included that spawns threads to start elections
while the tablets are being deleted. This test pretty reliably fails
without the included fix, and is solidly stable with the fix.
Change-Id: I4a0ea46f220a10d4de680fc73d45d14426e0395e
---
M src/kudu/integration-tests/delete_tablet-itest.cc
M src/kudu/tserver/ts_tablet_manager.cc
2 files changed, 59 insertions(+), 5 deletions(-)
git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/83/7883/2
--
To view, visit http://gerrit.cloudera.org:8080/7883
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings
Gerrit-MessageType: newpatchset
Gerrit-Change-Id: I4a0ea46f220a10d4de680fc73d45d14426e0395e
Gerrit-PatchSet: 2
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Mike Percy <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <[email protected]>
