Email Getting Stuck in Active Queue

[soxfan]

I am currently running Zimbra OSE version 4.0.3 on a CentOS 4 (RHEL 4) server. I am running in a split domain mode, as I am evaluating Zimbra as a possible replacement for our current POP3-only Linux-based email server. Right now I only have two active accounts on the Zimbra server; my own and a 'spam dump' account, which gets email tagged as spam by our email gateway server. The last couple of days I have been running into an issue where email messages are getting stuck in the Active mail queue. Today the queue got up to over 800 messages, which had been accumulating for over an hour, before I finally stopped and restarted Zimbra and cleared out a bunch of messages to the 'spam' account, as I don't really care about these. Looking around the system for clues on what is causing this problem, I came across errors like the follow in the /var/log/messages file:

Nov 7 14:23:59 email amavis[1803]: (01803-04-7) (!!) TROUBLE in check_mail: spam_scan FAILED: timed out
Nov 7 14:23:59 email amavis[1803]: (01803-04-7) (!) PRESERVING EVIDENCE in /opt/zimbra/amavisd/tmp/amavis-20061107T114500-01803

The time on the first message I see like this pretty much coincides with the time I start noticing things stuck in the queue. Once the first error is logged I see the same type of error messages repeating in the log. I don't really know if this error is indicative of the cause of the problem, or perhaps they are a result of some other issue. I'm not seeing any other error messages, although maybe I just don't know where to look.

I know the 'spam' account is receiving a very high volume of email messages, and this may be part of the problem. However, I am concerned about why the large volume would cause this backup of messages. If we decide to go live with Zimbra we will have between 300-400 active accounts, and this type of issue would not be good.

Any help on solving this problem would be greatly appeciated.

TIA,
John

[KevinH]

What is the CPU and MEM usage like? What do you have for disks on the zimbra partitions?

[soxfan]

Well, I've since fixed the problem by changing the way I handled spam at my email gateway, so I can't give you any exact numbers. When the problem was happening I was monitoring with 'top' and didn't see any process that was consistently taking up either CPU or MEM.

[soxfan]

I guess I'm not totally out of the woods on this yet. I'm not seeing the pile of messages stuck in the Active queue, however I'm still getting messages going to the deferred queue. The reason for the messages going to the deferred queue shows as "spam_scan FAILED: timed out (in reply to end of DATA command)". What does this error mean? Is it a network problem?

[KevinH]

Most of the time that means it timed out talking to SA for spam scanning.

[soxfan]

Any tuning or config changes I can make to prevent this from happening? Since I fixed my original problem the volume of messages is pretty low. So I don't think the server is being overwhelmed or anything like that.

[zagman76]

I am having a similar issue - where e-mail is being stuck in the active queue. The only resolution I have been able to come up with is running zmcontrol stop and start at least twice a day.

The software is running on FC4 w/ 2x1.8 ghz XEON Processors, 2GB RAM, 1-36GB (stand alone) drive for OS
1-72GB (stand alone) drive for zimbra & mailstore

Why is this happening? We only have about 200 users on the system - all using pop. Let me know if there is more you need from me.

Thanks!

[zagman76]

Has there been any other info on this?

I have another thread that started out as one issue, and migrated into this exact issue.

'Blackhole' e-mails send to non-zimbra accounts?

Does anyone know why this is happening, and how to correct it?

[rdunn]

Upgrade to 4.0.3 and mail delivery slows to a crawl
This seems to have fixed our server. I stopped zimbra and ran yum update, after it finished I rebooted and started Zimbra. The queue emptied within the hour and has been fine since (12 hours).

[phoenix]

Originally Posted by zagman76

Has there been any other info on this?

I have another thread that started out as one issue, and migrated into this exact issue.

'Blackhole' e-mails send to non-zimbra accounts?

Does anyone know why this is happening, and how to correct it?

Did you upgrade to the latest Zimbra release? Did you consider changing to another operating system such as CentOS?