[
https://issues.apache.org/jira/browse/CURATOR-126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14076733#comment-14076733
]
ASF GitHub Bot commented on CURATOR-126:
----------------------------------------
Github user dragonsinth commented on a diff in the pull request:
https://github.com/apache/curator/pull/23#discussion_r15487030
--- Diff:
curator-framework/src/main/java/org/apache/curator/framework/imps/CuratorFrameworkImpl.java
---
@@ -74,7 +71,7 @@
private final NamespaceFacadeCache namespaceFacadeCache;
private final NamespaceWatcherMap namespaceWatcherMap = new
NamespaceWatcherMap(this);
- private volatile ExecutorService executorService;
+ private volatile Thread backgroundThread;
--- End diff --
The bug is simple: Executor.shutdownNow() calls FutureTask.cancel(), which
on older JDK versions has a bug where it can interrupt the wrong thread,
causing all sorts of problems. Manually interrupting a thread does not have
this problem. Given that this Executor is only every used to run one task (the
backgroundOperationsLoop) we can simply switch to a Thread and avoid that bug.
> IllegalStateException in performBackgroundOperation during close
> ----------------------------------------------------------------
>
> Key: CURATOR-126
> URL: https://issues.apache.org/jira/browse/CURATOR-126
> Project: Apache Curator
> Issue Type: Bug
> Components: Framework
> Affects Versions: 2.5.0
> Reporter: Scott Blum
> Assignee: Cameron McKenzie
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> {code}
> [CuratorFramework-0] ERROR
> org.apache.curator.framework.imps.CuratorFrameworkImpl - Background
> exception was not retry-able or retry gave up
> java.lang.IllegalStateException: Client is not started
> at
> com.google.common.base.Preconditions.checkState(Preconditions.java:176)
> at
> org.apache.curator.CuratorZookeeperClient.getZooKeeper(CuratorZookeeperClient.java:113)
> at
> org.apache.curator.framework.imps.CuratorFrameworkImpl.performBackgroundOperation(CuratorFrameworkImpl.java:807)
> at
> org.apache.curator.framework.imps.CuratorFrameworkImpl.backgroundOperationsLoop(CuratorFrameworkImpl.java:793)
> at
> org.apache.curator.framework.imps.CuratorFrameworkImpl.access$400(CuratorFrameworkImpl.java:57)
> at
> org.apache.curator.framework.imps.CuratorFrameworkImpl$4.call(CuratorFrameworkImpl.java:275)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
> at java.lang.Thread.run(Thread.java:744)
> {code}
> I see this sometimes during test runs; I believe this happens because
> CuratorZookeeperClient.started gets set to false during shutdown, but the
> backgroundOperation loop can still be running since shutting down the
> backgroundOperation loop is inherently racy.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
