Zimbra becomes unresponsive for 5-40 minutes at a time

[pjsmith]

Hi,

Our zimbra server keeps going unresponsive. We've upgrade to the latest version a few days ago, but this problem has been there for about 3 versions now. About 2-3 times a day now it simply stop responding for anything between 5-40 minutes. Normally about 10 minutes or so, but just now 46 minutes. I can't see anything in the logs, and apart from zimbra going off line, everything else on the server seems to be OK. SMTP, POP, IMAP, and the web inteface become unresponsive. I can still log into the box OK etc. It comes back by itself after it's little sleep.

The box is not heavily loaded.

Any ideas where I start looking for issues?

I'm going to monitor logs and load carefully now, but was wondering if this might be a common issue or anyone has any pointers to what might be causing it.

[BrianA]

sounds a bit like what I've been getting. check /opt/zimbra/log/mailbox.log, I've got full of exceptions in that one.

[uxbod]

Perhaps some stats about the server ? How much load is present on the box ? Are you able to access the ZCS Admin GUI ? Anything else running on the server ?

[pjsmith]

Although this problem has been happening every day for quite some time and persisted through reboots and upgrades, it's now stopped!

I did a zmcontroll stop and start. When I logged back into the consoel it tsays the logger is not running. Without the logger service running it's been running smoothly for 24 hours+ now. Will run for longer and see if it stays this way.

[xeon]

Sounds like you had a lot of disk IO causing the unresponsiveness. This is very common on virtual environments that does not have good enough disk IO hardware.

[guises]

Sounds like you had a lot of disk IO causing the unresponsiveness. This is very common on virtual environments that does not have good enough disk IO hardware.

That doesn't sound right. My server covers a small office, with generally no more than a few hundred emails/hour, does nothing but email, and is brand new and I'm having this problem as well.

I'm looking forward to your update pjsmith - I need to resolve this somehow.

[phoenix]

Originally Posted by guises

That doesn't sound right. My server covers a small office, with generally no more than a few hundred emails/hour, does nothing but email, and is brand new and I'm having this problem as well.

I'm looking forward to your update pjsmith - I need to resolve this somehow.

How about giving some more information about the problem such as: hardware specifications (especially RAM); some diagnostic information from top (for instance) or zmstat-chart; which operating system and Zimbra version and release; how often does this happen; when did it start?

[pjsmith]

well, still hs no hiccups with the logger turned off. I'll try to get some stats, but as the logger isn't really system critical anyway, I'd say if you have this problem try turning the logger off and see if the issue goes away.

It is a virtual machine, not that it really matters. CentOS5, latest stable zimbra release, though as I said before, this issue has persisted through about the last 3 updates. It is on a huge server with more than enough disk io and monitoring the zimbra server from the guest so never shows more than a few percent or so waiting for disk. It has 1Gb ram. It's only a small box.

I'll update again once I have had time to investigate further, probably over the weekend

[xeon]

Originally Posted by pjsmith

well, still hs no hiccups with the logger turned off. I'll try to get some stats, but as the logger isn't really system critical anyway, I'd say if you have this problem try turning the logger off and see if the issue goes away.

It is a virtual machine, not that it really matters. CentOS5, latest stable zimbra release, though as I said before, this issue has persisted through about the last 3 updates. It is on a huge server with more than enough disk io and monitoring the zimbra server from the guest so never shows more than a few percent or so waiting for disk. It has 1Gb ram. It's only a small box.

I'll update again once I have had time to investigate further, probably over the weekend

I am still very curious...

Its a VM with 1GB of RAM on "huge server." 1GB of RAM on the VM is a little low. I have done it but it only "works" with the logger disabled (as you can see). Give it some more RAM that should help the Java processes out greatly.

What Virtual host software are you using?
How many VMs are running on the "huge server?"
How much RAM does your "huge server" have?
What is your disk hardware and RAID level?

[pjsmith]

I used to run it with less than 1g ram just fine. I'm using ESX and the box has 16Gb of ram running raid 10 with hardware raid and a 512Mb write cache. It flies. Far quicker than a nomal pc/server in terms of IO. 1Gb of ram for me on a linux server is a lot. The zimbra box is doing nothing else. I generally only give 512Mb to my other linux boxes It doesn't look like it's starved of ram and when the zimbra processes are hung other processes on the linux box seem just fine and the box does not appear lodaded. It's not the box that grinds to a halt, just zimbra... I'll see if I can get some more info over the weekend.