queue problem w/ deferred messages

[alderion]

Hi all. I installed zimbra a couple weeks ago and the install went very well. I enjoyed zimbra for a couple days and then started noticing that messages were pilling up in the deferred queue and I couldn't figure out why. I am getting messages in the log like the following:

Sep 26 04:07:45 mythtv postfix/lmtp[9754]: 6335B16EBAE: to=<pta-amidala@psaconsultants.com>, relay=none, delay=300, status=deferred (connect to mythtv.psaconsultants.com[67.165.234.150]: read timeout)

The problem is I can't figure out what that means. To solve the problem temporarily, I restart zimbra and then flush the queue and all the messages deliver. Everything works fine for a while and then things start to pile up in the deferred queue again.

I've searched the forums and it looks like there are some issues with permissions and the queue so I tried to run zmfixperms but it didn't fix the problem. I've also read that the system has to be up to date but it doesn't mention which packages. I've tried to run yum upgrade on the system but it fails due to dependecies so if I knew which packages to upgrade, I could try that.

Anyway, any and all pointers and help would be appreciated. My setup is:

Fedora Core 5 (slightly out of date)
Zimbra 4.0.1_GA_324

I will be happy to provide any other information/logs/etc.

Thanks,

-peter

[phoenix]

How do you have a GA version on FC5 as it only went GA in version 4..0.2? Try installing the new 4.0.2 for FC5 then run fixperms and see how you get on with that, post any problems you encounter.

[dijichi2]

psaconsultants.com is an external domain not related to you, i take it? the log message would imply some sort of network problem, like a firewall or smtp application layer firewall, or router issue. if a flush solves the problem its possible you have some sort of transient network issue? does this happen to all your external mail, and is your internal mail ok?

[alderion]

phoenix, i just pulled the version off of the rpm's that are installed. i thought i had downloaded the latest version that was available for fc5. perhaps the rpm's still have the old names. i'm at work at the moment and will check it out and then post results when i get home this evening.

dijichi2, psaconsultants.com is my main and only domain. i believe the message below pertains to an email that is being routed through the system via spam/av, etc. it does bring one thing to mind though. is there a way i can set zimbra up to talk to itself on 127.0.0.1? i ask becuase the ip shown is my one network routeable ip and is assigned to my router on which port 25 is then forwarded to the zimbra server but the host name is the zimbra server itself which has a nat address on the 192.168.0.0 network or, of course, 127.0.0.1.

thanks for the help.

-peter

[alderion]

nevermind, i went to the zcs sf homepage and then clicked on download and found the blessed version of fc5 binary. i'll try to install that. can i just untar it and run install.sh?

someone should update the link on the open source download page to point to the blessed packages. i'm sure it used to point to the right spot before zimbra added support for fc5.

ok, i went back to check what i downloaded and installed. it seems that the fc5 link on the open source edition download page goes to some developer builds. the only version available under 4.0.2 is zcs-4.0.2_GA_362.UBUNTU6.tgz which looks like an ubuntu package so went down to the 4.0.1 version where the only available package is zcs-4.0.1_GA_324.FC5.tgz. from the looks of it, that is the latest fc5 package so i installed that.

should i be downloading from somewhere else? is the fc5 link on the open source download page pointing to the wrong place? where can i get a 4.0.2 version of zimbra for fc5?

thanks!

-peter

[phoenix]

Yes, there seems to be an incorrect link on the download page. Go to the downloads page and click on the fc5 version, when you get to the sourceforge page just click on the 'file's entry near the top of the page and you'll see a full list of versions including the fc5 GA version, download and install that.

[alderion]

i upgraded to the blessed 4.0.2 version for fc5 and i've watched the queue for about 3 days and haven't seen even 1 deferred message. problem solved. i also notice that the link on the download page is fixed as well. thanks for the help. good product, good help.

-peter

[phoenix]

Glad your system is working. Yes, the link problem was just of links pointing to the wrong place.

[alderion]

ok, this has happened again. the server runs fine for a while and then stops delivering mail, sending everything to the deferred queue. after a zimbra restart and flush, the messages get delivered like normal.

one wrinkle is that the disk did fill up at one point due to a runaway process. i removed the huge file, brought down zimbra, did a fixperms and then brought zimbra back up but the issue came up again.

my question is, would the disk filling up cause this issue? if so, how can i fix it permanently? if not, how can i help you help me? :-) the logs don't seem very helpful, all they say is read timeout, message deferred.

one other thing i noticed was that there is a spam message with a ton of recipients in the queue and it almost looks like zimbra chokes on that because the last time i flushed, i had 33 deferred messages and 30 of them delivered and 3 didn't. i'm going to try and flush the 2 that don't look like spam to see if they deliver. could this be the cause of my problem? if so, how do i fix it as i've had my email address for years and i get a ton of spam.

on a related note, one of the emails that is being deferred is important. is there a way to get it off the disk out of the queue somehow?

my setup:

fc5
4.0.2_GA_362.FC5

thanks,

-peter

[Risotto]

I have the exact same problem here on a RHEL4 server with a networked edition. But we never had a disk fillup or not enough memory. This happened to us today while it happened back in late September too. It never occurred with versions lower than 4.0.1. This is getting annoying because even though we monitor the server's services we are not being notified when the queue begins to grow because of deferred messages. By the way, we got the same error message: read timeout on the local public ip address.

EDIT: To solve the problem we are doing a restart of the services: zmcontrol stop, zmcontrol start. Then, after a few minutes it is possible to flush the queue with postqueue -f. Strangely this is not possible right after the restart. We had to wait a few minutes for the mailserver not to deferr all messages again. It would be great if someone at zimbra could clarify what's happening. We really don't know what was wrong, whether amavis was not accepting connections anymore or mysql was down. These, by the way would be my two guesses, but I wasn't able to confirm this.