You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@curator.apache.org by "Michael Morello (JIRA)" <ji...@apache.org> on 2013/08/03 19:49:48 UTC
[jira] [Updated] (CURATOR-45) LeaderSelector threw exception, but
still created ephemeral node, breaking everything
[ https://issues.apache.org/jira/browse/CURATOR-45?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Michael Morello updated CURATOR-45:
-----------------------------------
Attachment: CURATOR-45.patch
Here's a first draft, the patch uses FailedDeleteManager to ensure that the znode is deleted once the connection is back.
> LeaderSelector threw exception, but still created ephemeral node, breaking everything
> -------------------------------------------------------------------------------------
>
> Key: CURATOR-45
> URL: https://issues.apache.org/jira/browse/CURATOR-45
> Project: Apache Curator
> Issue Type: Bug
> Reporter: Shevek
> Assignee: Jordan Zimmerman
> Attachments: CURATOR-45.patch
>
>
> ZooKeeper hiccupped, and then this happened:
> 2013-06-19 02:23:35,561 DEBUG [LeaderSelector-1] com.netflix.curator.RetryLoop.takeException (RetryLoop.java:184) - Retry-able exception received
> org.apache.zookeeper.KeeperException$ConnectionLossException: KeeperErrorCode = ConnectionLoss for /[REMOVED]/election/_c_1ccdb2b9-7f9a-4570-9555-201c91ec2dcb-lock-
> at org.apache.zookeeper.KeeperException.create(KeeperException.java:99) ~[zookeeper-3.5.0.jar:3.5.0--1]
> at org.apache.zookeeper.KeeperException.create(KeeperException.java:51) ~[zookeeper-3.5.0.jar:3.5.0--1]
> at org.apache.zookeeper.ZooKeeper.create(ZooKeeper.java:876) ~[zookeeper-3.5.0.jar:3.5.0--1]
> at com.netflix.curator.framework.imps.CreateBuilderImpl$10.call(CreateBuilderImpl.java:625) ~[curator-framework-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.imps.CreateBuilderImpl$10.call(CreateBuilderImpl.java:609) ~[curator-framework-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.RetryLoop.callWithRetry(RetryLoop.java:106) [curator-client-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.imps.CreateBuilderImpl.pathInForeground(CreateBuilderImpl.java:605) [curator-framework-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.imps.CreateBuilderImpl.forPath(CreateBuilderImpl.java:428) [curator-framework-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.imps.CreateBuilderImpl.forPath(CreateBuilderImpl.java:41) [curator-framework-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.locks.LockInternals.attemptLock(LockInternals.java:218) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.locks.InterProcessMutex.internalLock(InterProcessMutex.java:218) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.locks.InterProcessMutex.acquire(InterProcessMutex.java:74) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.leader.LeaderSelector.doWork(LeaderSelector.java:314) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.leader.LeaderSelector.doWorkLoop(LeaderSelector.java:373) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.leader.LeaderSelector.access$100(LeaderSelector.java:46) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at com.netflix.curator.framework.recipes.leader.LeaderSelector$2.call(LeaderSelector.java:195) [curator-recipes-1.3.5-SNAPSHOT.jar:?]
> at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334) [?:1.6.0_27]
> at java.util.concurrent.FutureTask.run(FutureTask.java:166) [?:1.6.0_27]
> at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1146) [?:1.6.0_27]
> at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) [?:1.6.0_27]
> at java.lang.Thread.run(Thread.java:679) [?:1.6.0_27]
> However, the ephemeral node got created, and this hung leader election for this path.
> I'm investigating to work out where to put an extra guaranteed-delete. I see the case in LockInternals, which sometimes triggers to do this cleanup, but it didn't trigger in this case.
> You must really love our bugs by now.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira