river-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Greg Trasuk (JIRA)" <j...@apache.org>
Subject [jira] [Resolved] (RIVER-327) Fast logging loop due to unexpected multicast socket close in LookupDiscovery$AnnouncementListener
Date Wed, 11 Dec 2013 05:45:07 GMT

     [ https://issues.apache.org/jira/browse/RIVER-327?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Greg Trasuk resolved RIVER-327.
-------------------------------

    Resolution: Unresolved

Been a while, and versions have moved on...

> Fast logging loop due to unexpected multicast socket close in LookupDiscovery$AnnouncementListener
> --------------------------------------------------------------------------------------------------
>
>                 Key: RIVER-327
>                 URL: https://issues.apache.org/jira/browse/RIVER-327
>             Project: River
>          Issue Type: Bug
>          Components: net_jini_discovery
>    Affects Versions: jtsk_2.1
>         Environment: Win Server 2003, Sun Java 1.5.0_12
>            Reporter: Chris Dolan
>
> In extremely rare circumstances (root cause still unknown) we see the following log message
thousands of times per second:
> INFO: exception while listening for multicast announcements
> java.net.SocketException: socket closed
>             at java.net.PlainDatagramSocketImpl.peekData(Unknown Source)
>             at java.net.DatagramSocket.receive(662)
>             at net.jini.discovery.LookupDiscovery$AnnouncementListener.run(1187)
> I don't know how or why the multicast listener socket has been closed (perhaps an OS
network stack glitch and not Jini's fault) but the consequence is that the system gets an
OutOfMemory error trying to process the vast number of log messages.
> Proposed solutions (not mutually exclusive):
>  * catch SocketException and if socket.isClosed() then rebuild the socket
>  * put a Thread.sleep(...) in the catch
> Note: this is a slightly modified version of Jini 2.1.  I've compared the code we're
running to the River trunk and confirmed that there's no differences in this section of the
code.



--
This message was sent by Atlassian JIRA
(v6.1.4#6159)

Mime
View raw message