I filed SOLR-4191 because I was seeing some unusual behavior -
exceptions when accessing /admin/mbeans and /admin/ping. Initially I
thought it was happening while updates/commits were going on, but after
discovering the same thing happening in my production 3.5.0 servers as
well as branch_4x from svn, I eventually tracked it down to long ParNew
GC pauses - clients are timing out and disconnecting because of those
pauses. SOLR-4191 is now closed as not a problem.
I have now reduced my Eden size from about 4GB to under 256MB (max heap
is still 8GB), but the long pauses are still happening on my Solr 3.5.0
servers, and they are still ParNew collections.
Here's some information about my setup:
-Xms4096M -Xmx8192M -XX:MaxNewSize=256M -XX:+UseParNewGC
-XX:+UseConcMarkSweepGC -XX:+CMSParallelRemarkEnabled
-XX:+CMSIncrementalMode -XX:MaxTenuringThreshold=3 -verbose:gc
-Xloggc:log/gc.log -XX:+PrintGCDateStamps -XX:+PrintGCDetails
java version "1.6.0_38"
Java(TM) SE Runtime Environment (build 1.6.0_38-b05)
Java HotSpot(TM) 64-Bit Server VM (build 20.13-b02, mixed mode)
Linux idxa1 2.6.32-279.19.1.el6.centos.plus.x86_64 #1 SMP Wed Dec 19
06:20:23 UTC 2012 x86_64 x86_64 x86_64 GNU/Linux
[root@idxa1 log]# cat /sys/block/sda/queue/scheduler
noop anticipatory [deadline] cfq
total used free shared buffers cached
Mem: 64934416 49360316 15574100 0 269580 38186012
-/+ buffers/cache: 10904724 54029692
Swap: 8388600 10560 8378040
CPU: two Intel E5440 per server, ranging from 2.5 to 2.83 Ghz.
What other information do I need to provide to track down what's happening?
Thanks,
Shawn
