[
https://issues.apache.org/jira/browse/TWILL-129?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14556603#comment-14556603
]
ASF GitHub Bot commented on TWILL-129:
--------------------------------------
Github user ghelmling commented on a diff in the pull request:
https://github.com/apache/incubator-twill/pull/34#discussion_r30920964
--- Diff:
twill-zookeeper/src/main/java/org/apache/twill/internal/zookeeper/LeaderElection.java
---
@@ -240,10 +240,10 @@ private void becomeFollower() {
* Starts watching for the max. of smaller node.
*/
private void watchNode(final String nodePath, Watcher watcher) {
- OperationFuture<Stat> watchFuture = zkClient.exists(nodePath, watcher);
- Futures.addCallback(watchFuture, new FutureCallback<Stat>() {
+ OperationFuture<NodeData> watchFuture = zkClient.getData(nodePath,
watcher);
--- End diff --
So this is relying on the fact that a KeeperException.NoNodeException will
be thrown if the node has already been removed? That would trigger the
onFailure() hook and re-run the election. Is that right?
> Race condition in leader election if leader quitted before follower set the
> watch
> ---------------------------------------------------------------------------------
>
> Key: TWILL-129
> URL: https://issues.apache.org/jira/browse/TWILL-129
> Project: Apache Twill
> Issue Type: Bug
> Components: zookeeper
> Affects Versions: 0.4.0-incubating, 0.5.0-incubating, 0.4.1-incubating
> Reporter: Terence Yim
> Assignee: Terence Yim
> Priority: Blocker
>
> In the LeaderElection implementation, if the leader quit and delete its node
> before the next-in-line follower try to set the watch through the exists()
> call, the follower will fail silently without noticing the leader node is
> gone.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
