Memory Leak?

[mtlhead089]

I spent this weekend trying to install to no avail. The system that I am using is as follows;

A new Dell PowerEdge 2950.
2 DualCore 3.0 Ghz Processors
4GB RAM
Gobs of harddrive space
RHEL 4

I am trying to install the latest version 4.5.6 GA 1024. I disable sendmail and apache from starting, I have moved ssh to a non-standard port. On install, everything seems to have gone well. On reboot, it seems to have gone well also but....

At a console (I am not running X, but init 3) I su - zimbra and do a 'zmcontrol status' I wait for approx 5 minutes and then that is it for the os. The console becomes unusable due to the all the 'Out of memory' errors and memory dumps. Last night I ran through a couple of reinstalls of the os just to make sure. This has been done with a plain install of RHEL4 with no updates. I have just completed a reinstall of the os and am currently in the process of getting all the latest updates. Then, I will try installing Zimbra NE again. Any thoughts, advice, or flames on the install for RHEL4 would be appreciated.

I'm tired. Been at this for too long. I missed my window to get this done this weekend to retire the old exchange server. I guess I'll spend my time this week on the install and ensuring it is stable before migrating.

[phoenix]

There should be no reason why there's a memory leak. I'd suggest putting ssh back on the standard port, it doesn't make any difference moving it to a non-standard port and Zimbra uses that port.

Is this a 64bit or 32bit install? Have you checked all the prerequisites for the install such as DNS A & MX records /etc/hosts file, compat libs etc.?

[mtlhead089]

I am very frustrated as I have not once been able to use this. This is very bad and I cannot understand what is going on because I know sysadmins are using Zimbra successfully. I have now tried this on 2 servers. Both are identical brand new Dell PowerEdge 2950's. After installing Zimbra, these servers are boat anchors. without they are just fine. I have spent the last couple of daze reinstalling RHEL4 first just the old 32 and then the 64 bit version just be sure I have the same problem. I might as well just setup sendmail/squirrelmail/clamav/spamassassin as I have done in the past. But, I want the additional features that Zimbra has put into their package.

[jholder]

Can you provide any type of diagnostic info to suggest that Zimbra has a memory leak?

x64 is more memory intensive that x86

[phoenix]

Originally Posted by mtlhead089

I am very frustrated as I have not once been able to use this. This is very bad and I cannot understand what is going on because I know sysadmins are using Zimbra successfully. I have now tried this on 2 servers. Both are identical brand new Dell PowerEdge 2950's. After installing Zimbra, these servers are boat anchors. without they are just fine. I have spent the last couple of daze reinstalling RHEL4 first just the old 32 and then the 64 bit version just be sure I have the same problem. I might as well just setup sendmail/squirrelmail/clamav/spamassassin as I have done in the past. But, I want the additional features that Zimbra has put into their package.

I did ask you if you had checked the DNS and hosts file prerequisites.

How about some info from the install log, is there any errors in there? Have you also disabled SElinux and the RHEL firewall? How about some evidence that there's a memory leak? Why do you think there is and what evidence is there for that?

It's a doddle to install on RHEL.

[mtlhead089]

Ok... No, I don't have any objective information. Just my subjective experience over the last few days. My frustration had boiled over and clouded my thoughts. So, I am going to begin again w/ a clear slate. I am going to start from a fresh install of RHEL4 x86. On my previous attempts I have downloaded the latest release on my laptop and burned it onto a cd w/ the license file. This time, I will directly download it to the server. I will transfer the license file via something else later (floppy, flashdrive, etc). I am going to document all symptoms and capture any logs as they appear.

[mtlhead089]

Originally Posted by phoenix

I did ask you if you had checked the DNS and hosts file prerequisites.

How about some info from the install log, is there any errors in there? Have you also disabled SElinux and the RHEL firewall? How about some evidence that there's a memory leak? Why do you think there is and what evidence is there for that?

It's a doddle to install on RHEL.

Yes, the hosts and dns prereqs were done and was fully accessible to the network in both directions (both by ip and fqdn). Yes; both the firewall and SELinux were disabled. I could not gather information due to the console unusable due to the oom error mesgs and mem dumps to the console. Rather than just reboot into single mode and review the logs I just reinstalled the os and retried to install zcs.

The server is currently in process of reinstall of the RHEL4 x86. I will gather subjective info if this occurs again.

[Verbunk]

Hi guys,

I apologize if I hijacked this thread but I'm kinda having the same problem. I'll go to login to check my mail (I'm the only user) I find FF unable to connect. A quick trip to the server finds a kernel crash due to OOM issues. It seems Java is the culprit.

I'm running 5.0B1 856.

I actually don't know how to find out exactly how/why this is happening but this machine runs ubuntu and Zimbra and that's it. I recently updated to 5B1 and this problem occured after the upgrade so it wasn't there from the start. The memory leak has gotten so bad I need to restart the server daily.

Any tips to find out what's wrong?

-J

[brian]

You really need to determine what is using the memory, use tools like ps and top while the problem is occuring.

In some cases I've seen problems with /bin/unicode_start getting into an endless loop and consuming all the memory on a RHEL system when starting applications from the console. A ps on a system that has this problem will show hundreds of these processes.

Changing #!/bin/bash to #!/bin/bash --norc fixes avoids the loop and fixes the problem in /bin/unicode_start on some systems.

[Verbunk]

Thanks Brian,

Using ps -aux shows /opt/zimbra/jdk1.6.0/bin/java is the culprit. Well, after a fresh reboot it's consuming 34% of my ram (384MB). I don't know if that's normal or not. Like I mentioned, I'm the only user and I solely use the web portal to view mail. I'll post again in a few hours on how much mem it's consuming at that point.

Here are the options of that process

-Xms303m -Xmx303m
-client -XX:NewRatio=2 -Djava.awt.headless=true
-Djava.io.tmpdir=/opt/zimbra/mailboxd/temp
-Djava.library.path=/opt/zimbra/lib
-Djava.endorsed.dirs=/opt/zimbra/mailboxd/common/endorsed
-Dzimbra.config=/opt/zimbra/conf/localconfig.xml
-Djetty.home=/opt/zimbra/mailboxd
-DSTART=/opt/zimbra/mailboxd/etc/start.config -jar
/opt/zimbra/mailboxd/start.jar zimbra

-J