I investigated the problem further. I dumped the output of netstat before and after I restarted sogo. Before the restart (running 12 h) I had 280 conns (steadily increasing all day), after the restart I was back to 40.

The diff shows that almost all of the persistent connections, which only go away by restarting sogo are somehow *ldap* related: (VERBUNDEN = ESTABLISHED):

< tcp 0 0 localhost.localdom:ldap localhost.localdo:53381 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:51077 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:54146 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:34959 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:50318 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:51081 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:34964 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:36243 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:41362 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:36240 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:49309 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:56988 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:44452 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:44451 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:39330 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:49312 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:46261 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:47551 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:60351 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:41661 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:57017 VERBUNDEN < tcp 0 0 localhost.localdom:ldap localhost.localdo:46264 VERBUNDEN
...


These ldap conns are used to authenticate any sogo related http/webdav request with the local ldap server, right? Why do they stay?

Is it possible to tell ldap to release these connections after a timeout?
Is it possible to switch to a socket-based-communictaion?


Thanks for your time
John






On 01/17/2013 12:04 PM, John Bieling wrote:
My sogo installation is running on a virtuozo system (server4you.com) and I have a limit of 1200 "numtcpsock". If I sync my adressbook via thunderbird, I can see via

cat /proc/user_beancounters

that the number of held numtcpsock is going up. But it is not going down again, it just stays there. Now if my server runs for about a day, I reach the limit, which will "disconnect" my server from the outside, since no new connections can be established.

Before reaching the limit, I can ssh into the maschine and restart sogo and the "numtcpsock" value drops from about 1000 back to 50 (which is normal). This is now automated by a cron job, but why is that happening?

Has someone else experienced this?

In my sogo profile I have setup an extra IMAP account (gmx.de) and when I click through the gmx folders in the sogo web-interface, for each folder-click the "numtcpsock" value increases by 1 or 2 as well. And is not going back down.

Any Idea? The incomming connections vor CardDAV and CalDAV are http request, right? Can I tweak something there?

Thanks
John

--
[email protected]
https://inverse.ca/sogo/lists

Reply via email to