Imap connections

Running out of imapd connections

The problem looks like this from xymon:

yellow  Wed Feb  6 11:02:18 2013 imap NOT ok 

Service imap on localhost is not OK : Unexpected service response


Seconds: 0.000132

And clients see intermittent errors trying to connect.

And syslog shows this:

Feb  6 11:18:20 weasel imapd: 40 maximum active connections.
Feb  6 11:18:20 weasel imapd: Maximum connection limit reached for ::ffff:10.0.0.66
Feb  6 11:18:23 weasel imapd: last message repeated 3 times
Feb  6 11:18:28 weasel imapd: Maximum connection limit reached for ::ffff:10.0.0.66
Feb  6 11:18:28 weasel imapd: Maximum connection limit reached for ::ffff:10.0.0.66

This shows which clients are holding onto connections:

root@weasel:/etc/init.d# netstat -n |grep :143 | sort -k +4

Count which ones are in the ESTABLISHED state:

netstat -n |grep :143 | grep ESTAB |  sort -k +4 | wc

When my phone is connecting through the telco network instead of wifi, the connection looks like this: Verizon (myvzw.com)

tcp6       0    242 216.99.216.99:143       70.199.228.147:11752    ESTABLISHED

T-Mobile (tmodns.net)

tcp6       0      0 216.99.216.99:143       208.54.32.249:64691     ESTABLISHED

This shows the processes:

root@weasel:/etc/init.d# ps -eaf | grep imapd | wc
     39     397    3487

Why?

I know we have a lot of devices laying around. But what changed? Did the clients change or did the server change?

Raise the limits

http://kb.mediatemple.net/questions/259/How+can+I+raise+Courier-IMAP+connections+for+my+server%3F

Immediately after raising the limit and restarting imap, I had 13 active connections, all in the established state. A nexus tablet had the most connections open, 4. The other clients had one or two.