You are viewing a plain text version of this content. The canonical link for it is here.

Posted to mapreduce-user@hadoop.apache.org by Dmitry Simonov <di...@gmail.com> on 2015/03/03 14:37:11 UTC

DataNode: long GC times (page faults?)

Hello!

We are using HA Hadoop (v2.5.1) Cluster with 3 DataNodes on Windows Server
2008R2 with 24 GB RAM.
Sometimes we observe long GC times, e.g.:

1029537.175: [GC (Allocation Failure) 1029537.176: [ParNew:
5459023K->309990K(5662336K), 32.6609790 secs]
10328788K->5583905K(11953792K), 32.6611705 secs] [Times: user=0.92
sys=0.06, real=32.65 secs]

Parameters:
set GC_OPTS=-XX:+UseNUMA -XX:+UseParNewGC -XX:+UseConcMarkSweepGC
-XX:+CMSParallelRemarkEnabled -XX:ConcGCThreads=4 -XX:ParallelGCThreads=15
-Xmx12G -Xms12G -Xmn6G

java version "1.8.0_25"
Java(TM) SE Runtime Environment (build 1.8.0_25-b18)
Java HotSpot(TM) 64-Bit Server VM (build 25.25-b02, mixed mode)

Could you please give some advice, how to troubleshoot this issue?

Best regards, Dmitrii Simonov.