[
https://issues.apache.org/jira/browse/SPARK-2282?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14065121#comment-14065121
]
Aaron Davidson commented on SPARK-2282:
---------------------------------------
This problem does look identical. I think I gave you the wrong netstat command,
as "-l" only show listening sockets. Try with "-a" instead to see all open
connections to confirm this, but the rest of your symptoms align perfectly.
I did a little Googling around for your specific kernel version, and it turns
out [someone else|http://lists.openwall.net/netdev/2011/07/13/39] has had
success with tcp_tw_recycle on 2.6.32. Could you try to make absolutely sure
that the sysctl is taking effect? Perhaps you can try adding
"net.ipv4.tcp_tw_recycle = 1" to /etc/sysctl.conf and then running a "sysctl
-p" before restarting pyspark.
> PySpark crashes if too many tasks complete quickly
> --------------------------------------------------
>
> Key: SPARK-2282
> URL: https://issues.apache.org/jira/browse/SPARK-2282
> Project: Spark
> Issue Type: Bug
> Components: PySpark
> Affects Versions: 0.9.1, 1.0.0, 1.0.1
> Reporter: Aaron Davidson
> Assignee: Aaron Davidson
> Fix For: 0.9.2, 1.0.0, 1.0.1
>
>
> Upon every task completion, PythonAccumulatorParam constructs a new socket to
> the Accumulator server running inside the pyspark daemon. This can cause a
> buildup of used ephemeral ports from sockets in the TIME_WAIT termination
> stage, which will cause the SparkContext to crash if too many tasks complete
> too quickly. We ran into this bug with 17k tasks completing in 15 seconds.
> This bug can be fixed outside of Spark by ensuring these properties are set
> (on a linux server);
> echo "1" > /proc/sys/net/ipv4/tcp_tw_reuse
> echo "1" > /proc/sys/net/ipv4/tcp_tw_recycle
> or by adding the SO_REUSEADDR option to the Socket creation within Spark.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
