java.io.IOException: Too many open files !!!

[Samp]

This morning I got this error in mailbox.log :

2010-05-11 10:11:23,189 WARN [btpool0-3 - Acceptor0 SelectChannelConnector@0.0.0.0:80] [] log - EXCEPTION
java.io.IOException: Too many open files

Help !

Sam.

[phoenix]

Originally Posted by Samp

This morning I got this error in mailbox.log :

2010-05-11 10:11:23,189 WARN [btpool0-3 - Acceptor0 SelectChannelConnector@0.0.0.0:80] [] log - EXCEPTION
java.io.IOException: Too many open files

Help !

Several threads in the forums cover this issue, a search would be the best place to start: site:zimbra.com "Too many open files" - Yahoo! Search Results = try some of those solutions first.

[Samp]

Ok, but this morning it was the panic !

I restart several time and now it seems good.

I will follow your link to understand the origin of this problem...

Sam.

[Samp]

I always the zimbra crash problem with java "Too many open files" on vserver Debian squeeze/sid

I checked and there is the /proc/vrtual

vserver-1:~# cat /proc/virtual/40000/limit
Limit current min/max soft/hard hits
PROC: 356 0/ 406 -1/ -1 0
VM: 1406127 0/ 3379795 -1/ -1 0
VML: 0 0/ 0 -1/ -1 0
RSS: 509070 0/ 537138 -1/ -1 0
ANON: 454452 0/ 481445 -1/ -1 0
RMAP: 54618 0/ 59754 -1/ -1 0
FILES: 3907 0/ 4083 524288/ 524288 0
OFD: 2529 0/ 2551 -1/ -1 0
LOCKS: -99 0/ -1 -1/ -1 0
SOCK: 582 0/ 582 -1/ -1 0
MSGQ: 0 0/ 0 -1/ -1 0
SHM: 0 0/ 0 -1/ -1 0
SEMA: 1 0/ 1 -1/ -1 0
SEMS: 1 0/ 1 -1/ -1 0
DENT: 2687 0/ 2756 -1/ -1 0

As you see the open files limit is 524288 and current is 3907 and max 4083 but 2mn ago Zimbra crashed and malbox.log show many java "Too many open files" errors and vtop show more than 100% java CPU use.

I need help because I can't stay all the day on console to restart Zimbra.

Sam.

[iway]

How much memory do you have allocated to the VM?

[Samp]

Originally Posted by iway

How much memory do you have allocated to the VM?

The machine is 8Gb memory and I haven't set a VM limit for this vserver, but in fact I make a mistake by upgrade system from Debian Lenny to squeeze/sid and I think this cause the problem, in Lenny I haven't this problem before and Zimbra worked fine with the same number of users and requests.

I created a new vserver guest with a regular Lenny setup, and now the ulimit -n is set with the right value at start with /etc/vservers/<guest>/rlimits/nofile

I re-install Zimbra on this new vserver on Debian Lenny and I'm transferring my data from old to new vserver. I hope this will work, I copy the folders "index" "store" and "...apache/derby"

Thanks for your help.

Sam.