[
https://issues.apache.org/jira/browse/FLUME-3257?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
zhenzhao wang resolved FLUME-3257.
----------------------------------
Resolution: Fixed
Release Note: The problem is already resolved in 1.8.0 as libthrift is
upgraded to 0.9.3
> Tcp mem and socket domain socket leak in flume 1.7 and earlier versions.
> -------------------------------------------------------------------------
>
> Key: FLUME-3257
> URL: https://issues.apache.org/jira/browse/FLUME-3257
> Project: Flume
> Issue Type: Bug
> Affects Versions: 1.7.0
> Reporter: zhenzhao wang
> Priority: Major
> Fix For: 1.8.0
>
>
> We'd like to report a resource(tcp mem, domain socket) leak caused by
> THRIFT-2402 . The problem is solved in 1.8.0 as it upgraded libthrift to
> 0.9.3, however we think it's still useful to share our experience with flume
> users. The sources which use libthrift server might encounter the resource
> leak problem, e.g. ScribeSource. The bug could be very serious (leading node
> unreachable in our case) while satisfying following conditions:
> # Flume stop and start thrift server (e.g. scribe source) in the same
> process due to configuration reload and etc. Every stop and start action will
> cause resource leak.
> # The server's traffic is very high. The leak might be quite big when the
> traffic is high.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
