Zimbra hangs server - problem with slapd?

[paul_rogers6]

Hi all

I have a Zimbra server that has been running for about three months now. Yesterday on trying to collect mail from it there was no response from the web server at the normal address. Checking the server showed that it would not respond to either ssh (no route to host) or a ping (Destination unreachable). At this point perhaps I should mention the server is running as a Xen virtual machine. Trying to connect to the Xen console just gives a blank screen. Trying to reboot or shutdown the machine does not work and I was eventally forced to force the machine to shutdown (pull the virtual plug).

When the server is restarted it starts until the zimbra service has been starting for approximately two to three minutes and then the machine hangs again. Any connection to the console or ssh then locks up. Once again the machine is then completely unresponsive.

I have taken the zimbra service out of the various run levels and the machine starts and stays up overnight but as soon as the zimbra service is started the machine locks up again.

I have tried starting the ldap service on its own (since this is the first service to start) and this gives exactly the same result - locking up the machine.

I tried this this morning and then restarted the machine first thing this evening. /var/log/messages contains multiple occurrences of the following message:

Sep 13 08:34:05 mta001 kernel: 4gb seg fixup, process slapd (pid 2982), cs:ip 73
:009e4a40

and this is the last message before the machine hung and the next message show the machine restarting this evening.

I'm not sure if this is relevant or not.

Can anyone suggest in the first instance where to look to discover what is the problem? I am assuming the problem is with zimbra but don't know where to look to find out what the problem is. I have tried checking /opt/zimbra/log/startup.log but this doesn't exist once the machine has restarted.

Can anyone help? Where should I look first?

Many thanks

Paul

[phoenix]

Bug 39498 – Xen: 4gb seg fixup, process slapd immediately after install

[paul_rogers6]

Hey Bill

Many thanks for the reply. Had a look at the bug notice. Do you think it could be this that is causing the server to hang? Doesn't it seem odd that the server has been running for some three months before this error caused the hang?

Can you suggest a way to figure out exactly what is causing the problem, eg other logs that might show something?

Again, many thanks.

Regards

Paul

[paul_rogers6]

Hey Guys

I've discovered that while the zimbra service is not running something is still writing to zimbra.log. This contains the following error message:

Sep 14 07:10:34 mta001 zimbramon[9136]: 9136:info: zmmtaconfig: Skipping getAllMtaAuthURLs ERROR: service.FAILURE (system failure: ZimbraLdapContext) (cause: javax.naming.CommunicationException mta001.my.net:389)

Presumably this is because the service is not running? Or does it point to the problem with zimbra?

Many thanks

Regards

Paul

[paul_rogers6]

Hey All

I'm having problems with slapd and xen, specifically the "kernel: 4gb seg fixup, process slapd" issue.

I've read the various posts but what no one seems to mention is whether the only outcome of the error is a log full of messages or whether this can result in other problems (the server hanging in my case).

Can anyone who has experienced the issue tell me what the ultimate outcome of the problem is?

In my case the server has run with no problems for three months or more (with error messages in the logs) but now hangs shortly after the ldap service starts. I've checked all the logs and can find nothing to indicate what the problem is. Could the 4gb seg fixup cause this problem? Or should I be looking else where?

Many thanks

Regards

Paul

[phoenix]

I thought I'd answered this in your other thread with the bug details, if you read that you'd see that this is a problem with the 32bit xen kernel - there's currently no fix for it (AFAIK) and you should migrate to another platform or 64bit xen (where the problem doesn't exist).

[paul_rogers6]

Hey Bill

Thanks for the reply - can you confirm this is what is hanging the server? I guess by the tone of your reply you mean yes? Sorry for being so dense.

Thanks

Paul

[phoenix]

Originally Posted by paul_rogers6

Thanks for the reply - can you confirm this is what is hanging the server? I guess by the tone of your reply you mean yes?

The error message you posted would lead me to believe that's what is happening. The problem with that error is that Zimbra may work for a while and some users may never see the problem at all (I'm one of them as I ran my Zimbra server under 32bit xen for a couple of years without problems) but when you get the error it seems impossible to fix it - you could always ask on the xen mailing lists to see if it's been resolved but I doubt it.

If you have the ability to move to a 64bit platform then I'd suggest that would be the best course of action (the problem doesn't happen in 64bit xen) and in the not too distant future all binary releases of Zimbra will be 64bit only.

[paul_rogers6]

Hi Bill

Many thanks for the reply. Looks like I'm moving server :-)

All the best

Paul