java.lang.OutOfMemoryError: Java heap space

[igeorg]

Hi,

We're using Zimbra Beta M2 on RHEL 4 (kernel 2.6.9-11.EL)
Yesterday we saw the following in zimbra.log and the server stopped responding:

~100 lines like this:
2006-02-02 23:43:06,340 INFO [http-443-Processor90] [mid=20;ip=x.x.x.x;name=x@x.com;] FileBlobStore - deleting blob 17216 in mailbox 20

and then:
2006-02-02 23:43:19,302 FATAL [http-443-Processor90] [mid=20;ip=x.x.x.x;name=x@x.com;] system - handler exception
java.lang.OutOfMemoryError: Java heap space

Output from ps -ef|grep tomcat:
/opt/zimbra/jdk1.5.0_05/bin/java -Xms455m -Xmx455m -client -XX:NewRatio=2

When it crashed, the server was running for about a month. It has ~100 users & processes ~2000s messages/day.

The box has 1.5Gb of memory. Do you think, the heap size should be increased? Any reason to believe it's a memory leak?

Let me know if you need more info.

Regards
John

[KevinH]

Was a user doing a batch delete from the web ui? Or was this a cron job that may have fired like message aging? If your sure there was no user interaction can you post your cron tab. If there was user interaction can you try to find out what it was? Empty Junk? Trash?

[igeorg]

Yes, the user shown in the logs just before the OutOfMemoryError, has been doing junk clearance + trash delete of his mailbox at that time or near that time

Let me know if you need more info to debug this

John

[KevinH]

How much memory does your server have? Can you also run "ps -ef | grep tomcat" to get the java options. About how many or how large was the mail that this user was deleting?

[igeorg]

Output from ps -ef|grep tomcat:
/opt/zimbra/jdk1.5.0_05/bin/java -Xms455m -Xmx455m -client -XX:NewRatio=2

The box has 1.5Gb of memory

At the time zimbra logged the OutOfMemoryError, the user was in the process of deleting ~200 messages.

Perhaps there are some operations in the java source base which are linear with respect to their requirements for memory? In that case an increase in -Xmx would reduce the frequency of these errors.

[KevinH]

In general there shouldn't be that big of a memory jump just by deleting. In fact it may have been about out of memory and the delete put you over. Has this happened again? We'll do some tests around deleting and try to reproduce this.

[igeorg]

We had 2 other failures in the past 3 moths with the same version/environment - M2, but didnt seem related to an OutOfMemoryError

Thanks for your help with this. We'll keep you updated

[KevinH]

Thanks we also added some huge delete tests in our test harness. Deletes of 5000 messages so if there are any memory related issues we should catch them there.

[John Birdsey]

I am also experiencing the java.lang.OutOfMemoryError: Java heap space issue. I am running Fedora core 3 with 1Gb Ram and I am testing the network version. But the mailbox service will not stay up and keeps falling over. It has just failed with...

2006-03-07 11:05:43,080 INFO [LmtpServer-6] [] misc - # of shared deliv decr to 0 for mailbox 8
2006-03-07 11:05:44,532 FATAL [LmtpServer-8] [] system - Fatal error occurred while handling connection
java.lang.OutOfMemoryError: Java heap space
2006-03-07 11:05:44,981 WARN [Convertd-3] [] TransformationHandler - got EOF from client

Please let me know if you need any further information as I am willing to do any tests you require.

Thanks for a great product.

John

[marcmac]

If you're getting an out of memory error, could you tell us how much memory is on your box?

What kind of system was that, anyway?

And the OS?