https://issues.apache.org/bugzilla/show_bug.cgi?id=56213
Bug ID: 56213
Summary: Many KeyReference Objects Cause Long CMS-remark GC
Times
Product: Tomcat 7
Version: 7.0.47
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P2
Component: Connectors
Assignee: dev@tomcat.apache.org
Reporter: jmid...@gmail.com
This is a request to prop the following fix to Tomcat 7.
http://svn.apache.org/viewvc?diff_format=h&view=revision&revision=1388890
Using the NIO connector with a WebSocket application many (millions of)
KeyReference objects are created under traffic. While the objects are small,
the sheer number of them seems to overwhelm the CMS collector. Before the
above fix with our application the CMS remark stop the world phase was taking
up to 8.8 seconds.
716.463: [GC[YG occupancy: 213377 K (392896 K)]716.463: [Rescan (parallel) ,
5.7904380 secs]722.253: [weak refs processing, 2.3101000 secs] [1 CMS-remark:
1160677K(1703936K)] 1374054K(2096832K), 8.8005960 secs] [Times: user=107.17
sys=0.00, real=8.80 secs]
Total time for which application threads were stopped: 8.8020690 seconds
After applying the fix under the same traffic the time drops to 110ms.
1282.459: [GC[YG occupancy: 203760 K (392896 K)]1282.459: [Rescan (parallel) ,
0.0594610 secs]1282.519: [weak refs processing, 0.0452530 secs] [1 CMS-remark:
1159948K(1703936K)] 1363708K(2096832K), 0.1084940 secs] [Times: user=1.12
sys=0.00, real=0.11 secs]
Total time for which application threads were stopped: 0.1100650 seconds
For a real time application the 9 second pause is a killer.
These tests were executed under:
RHEL 6.4
java version "1.6.0_65"
Java(TM) SE Runtime Environment (build 1.6.0_65-b14-BR8013809)
Java HotSpot(TM) 64-Bit Server VM (build 20.65-b04, mixed mode)
With the following JVM parms:
-server
-XX:NewSize=384m
-XX:MaxNewSize=384m
-XX:+UseParNewGC
-XX:MaxTenuringThreshold=0
-Xms2048m
-Xmx2048m
-XX:+UseConcMarkSweepGC
-XX:CMSInitiatingOccupancyFraction=68
-XX:+UseCMSInitiatingOccupancyOnly
-XX:PermSize=96m
-XX:MaxPermSize=96m
-XX:+UseMembar
-XX:+HeapDumpOnOutOfMemoryError
-XX:+PrintClassHistogram
--
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
