LMTP server closing connection - system non-responsive

[deepblue]

Originally Posted by stuartg.orion

The release statement above implies it's related to IMAP connections,

Hmm. I really doubt that this is related to IMAP. The release statement you are referring to, is not a official Zimbra release statement as far as I can see.
And I am still unable to find a bugreport for the problem (I searched for LMTP but all bugs I found are unrelated).

I have attached 2 screenshots (from ZCS NE 7.1.1 p2). lmtp_conn.jpg shows that the number of LMTP processes is going to the max. /var/log/zimbra.log starts showing that Postfix can not connect to LMTP:

refused to talk to me: 421 zimbra.server.com Zimbra LMTP server closing connection; service busy

I think that happens when the number of LMTP Processes is maxed out and all running LMTP Processes are busy. Question is: why are the LMTP processes "hanging"?

The second screenshot imap_conn.jpg shows the current number of imap connections. The peeks are at the same time (the screenshots have different scaling).
I have seen much more imap connections to that server, so I doubt, that IMAP is really the cause of the problem...

My assumption is, that LMTP can not put the mail into the mailbox for some reason. That causes new LMTP processes to be spawned until the max. of 20 processes/threads.
The only way fix this is to restart the mailboxd.
However: I found no indication in the mailbox.log why LMTP can not deliver the mail.

Regards
Thomas

[chauvetp]

Having the same problem. Only happened twice over the last two weeks but one of them was just now (an extremely quiet time in terms of system load for us, 8-9PM local time).

As other have reported, mailboxd fixes the issue, but isn't the most pleasant fix. I have not yet upgraded to 7.1.2 (but being the beginning of the semester, now isn't the greatest time - maybe next week).

[sae65]

Opened a ticket with Zimbra today. They stated that mini-patch 5 for Zimbra 7.1.1 and 7.1.2 does fix the issue.

[bjquinn]

Just wanted to add that in my case I was having the same symptoms in the log but it didn't seem to have anything to do with the number of connections or length of time the server had been up. Turned out I had a stray copy of freshclam running even after stopping all zimbra services. So, after stopping all the zimbra services, I killed the stray freshclam process, restarted zimbra and all was well.