/tmp/ folder completely filled

[sjobeck]

Hey, Hi, All,

Had an odd occurrence the other afternoon that none of us has ever seen in all of our years working with Zimbra.

I googled on it & this was the only other thing I found but this is not quite what happened: /tmp/convert folder fill up disk space - Zimbra :: Forums

We are running about 1010 users on "Release 5.0.16_GA_2921.UBUNTU8_64 UBUNTU8_64 NETWORK edition" on a huge IBM server with plenty of resources. (yes, yes, I know, we are upgrading right now)

The box has historically been rock-solid & is dedicated only to Zimbra & nothing else.

Two afternoons ago, Nagios yelled that /tmp/ was full but there was basically nothing in /tmp/ that we could see.

We then ran 'df' shows 100% full, but 'du' was showing 400k actual use. It took `lsof | grep /tmp` to see which process was holding any large
files open. This looked fishy:

sh 16566 zimbra 1w REG 254,1 0
8 /tmp/logprocess.out
sh 16566 zimbra 2w REG 254,1 0
8 /tmp/logprocess.out
zmlogproc 16568 zimbra 1w REG 254,1 0
8 /tmp/logprocess.out
zmlogproc 16568 zimbra 2w REG 254,1 0
8 /tmp/logprocess.out


I switched to the zimbra user and ran `zmloggerctl restart`, which
released the files (which had since disappeared) and space returned.

Any one else ever seen this before?

Our users got some hideous error messages (ie: SOAP this & SOAP that & service this & service that, kinds of errors) from a slug of messages that bounced back to them at about this exact same time as we restarted this daemon, so we are not 100% certain it was from this but the timing sure was coincidental.