tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Didier Chaumond" <didier.chaum...@atchik.com>
Subject Catalina eat 100%CPU under heavy load
Date Tue, 23 Aug 2005 09:11:14 GMT
Hi all,

We are using catalina in production environment for one of our production
product/platform Apache/mod_jk and Ajp13Connector, average traffic is around
15 req/s (catalina entries).

Apache http server 1.3.x version, mod_jk and two load balanced catalina
workers. Tomcat version is 4.1.30, RH9/Linux version 2.4.20, java version
"1.4.2_03", Sun jvm build 1.4.2_03-b02. 

After some hours (days ... it depends), CPU load grow to 100% on one of our
catalina server and catalina server "slow" some hours after finally it
doesn't respond anymore (but it's better on our point of view as we have
some other daemon checkers programs). Since we have patch our product
including optimisations on our side (database application), we encouter this
problem very often, now one or two times a day. We have first change
Ajp13Connector settings min/maxProcessors but this doesn't solve the
problem.

The thread dump show a lot of Ajp13Processors not working, in "waiting"
state:

"Ajp13Processor[8097][8]" daemon prio=1 tid=0x8a772530 nid=0x35da in
Object.wait() [8ac7e000..8ac7ec98]
        at java.lang.Object.wait(Native Method)
        - waiting on <0x49a01d50> (a org.apache.ajp.tomcat4.Ajp13Processor)
        at java.lang.Object.wait(Object.java:429)
        at
org.apache.ajp.tomcat4.Ajp13Processor.await(Ajp13Processor.java:295)
        - locked <0x49a01d50> (a org.apache.ajp.tomcat4.Ajp13Processor)
        at
org.apache.ajp.tomcat4.Ajp13Processor.run(Ajp13Processor.java:567)
        at java.lang.Thread.run(Thread.java:534)

385 processors / 500 are in this state, others seems to "work" ok.

I've already found some posts which talk about problems with the Ajp13 pool
implementation under heavy load, some others related to JVM problems. We
have also note (on other application/java servers) that JVM may be the
problem under heavy load situations and "overloaded" synchronized code
implementation.

Have anybody encounter this kind of problem, what are the solutions we can
investigate ? Changing Connector implementation, jvm version, tomcat version
... Or simply changing servlet server :-) !



---------------------------------------------------------------------
To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: tomcat-user-help@jakarta.apache.org


Mime
View raw message