[
https://issues.apache.org/jira/browse/CASSANDRA-1438?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Ellis updated CASSANDRA-1438:
--------------------------------------
Attachment: 1438.txt
Patch based on Erik's from CASSANDRA-1766. Changes:
- Merged StreamOutManager.cancelPendingFiles with StreamOutManager.remove
- allow StreamingService.cancelOutboundStreams to propagate
UnknownHostException up so JMX caller gets feedback about the problem
- removes unused StreamOutManager.pendingDestinations
- Removed manual catch in FileStreamTask. Exceptions will be logged by the
executor (MessagingService.streamExecutor_ is a DebuggableThreadPoolExecutor
like almost all our executors, which will log uncaught exceptions from its tasks
- No finishAndStartNext on FST failure; if streaming fails we don't want to
pretend it's succeeded
- Similarly SOM.remove does not notify its condition, which would be taken to
mean success by waiting threads
Those last two illustrate the hard part here -- how do we un-jam those threads,
without having them treat it as success, which it's not?
> Stream*Manager doesn't clean up broken Streams
> ----------------------------------------------
>
> Key: CASSANDRA-1438
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1438
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Affects Versions: 0.7 beta 1
> Reporter: Nick Bailey
> Assignee: T Jake Luciani
> Fix For: 0.7.1
>
> Attachments: 1438.txt
>
>
> StreamInManager and StreamOutManager only remove stream contexts/managers
> when a stream completes successfully. Any broken streams will cause objects
> to hang around and never get garbage collected.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
