[
https://issues.apache.org/jira/browse/CASSANDRA-5037?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13609218#comment-13609218
]
Brandon Williams commented on CASSANDRA-5037:
---------------------------------------------
I just checked on 3 separate providers, as well as a few other machines I have
and none have a /tmp partition. So dumping to /tmp now gives the ability to
write, but also fill up the root filesystem. IMO, it's less harmful to say
"you need to set this to a sensible location and reproduce the dump" than "you
need to deal with all the problems a full root fs causes, then set this to a
sensible location and reproduce the dump."
> Set CASSANDRA_HEAPDUMP_DIR in cassandra-env.sh by default
> ---------------------------------------------------------
>
> Key: CASSANDRA-5037
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5037
> Project: Cassandra
> Issue Type: Bug
> Components: Config
> Affects Versions: 1.1.7, 1.2.0 beta 3
> Reporter: J.B. Langston
> Priority: Minor
>
> Cassandra is configured via -XX:+HeapDumpOnOutOfMemoryError to trigger a heap
> dump at the occurrence of an OutOfMemoryError, but unless you have set your
> CASSANDRA_HEAPDUMP_DIR in cassandra-env.sh, the file will be written in the
> Cassandra process's working directory, which when Cassandra is run as a
> service is typically the root directory. Nine times out of ten, the Cassandra
> process does not have permission to write here so no heap dump is created.
> Even if Cassandra does have permission, the root filesystem is usually small
> and the heap dump could easily fill it up with a large Xmx configured. This
> makes post-mortem analysis difficult. We should set it by default to e.g.
> /var/log/cassandra.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
