[
https://issues.apache.org/jira/browse/STORM-1976?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15381359#comment-15381359
]
Jungtaek Lim commented on STORM-1976:
-------------------------------------
I didn't mean configuration of replication count. I'm saying current replicated
count for specific blob.
Anyway I'm just thinking about getting rid of that code, since cleaning up
corrupt topology when only startup is not making sense for now. We could
eventually set stable condition for checking corrupted topology, and clean up
periodically, but for now I don't have an idea.
> Storm Nimbus H/A has issue on cleaning corrupted topologies
> -----------------------------------------------------------
>
> Key: STORM-1976
> URL: https://issues.apache.org/jira/browse/STORM-1976
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-core
> Affects Versions: 1.0.0, 1.0.1
> Reporter: Raghav Kumar Gautam
> Assignee: Jungtaek Lim
> Priority: Blocker
>
> In the following scenario storm-ha runs into issues:
> 1. Kill a non-leader nimbus
> 2. Submit a topology
> 3. Bring up the non-leader nimbus
> After step-3 expectation is that the non-leader nimbus will download topology
> jar. Instead it cleans up the topology.
> {code}
> 2016-07-12 07:11:09.511 o.a.s.c.zookeeper-state-factory [WARN] Received event
> ::none: with disconnected Reader Zookeeper.
> 2016-07-12 07:11:09.587 o.a.s.zookeeper [INFO] Queued up for leader lock.
> 2016-07-12 07:11:09.608 o.a.s.d.nimbus [INFO] Corrupt topology
> JoinedNonLeaderNimbusTriesToDownloadTopologyCode-2-1468307239 has state on
> zookeeper but doesn't have a local dir on Nimbus. Cleaning up...
> 2016-07-12 07:11:09.932 o.a.h.m.s.s.StormTimelineMetricsReporter [INFO]
> Preparing Storm Metrics Reporter
> 2016-07-12 07:11:09.946 o.a.s.d.m.MetricsUtils [INFO] Using statistics
> reporter
> plugin:org.apache.storm.daemon.metrics.reporters.JmxPreparableReporter
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
