Hello ! I've installed NAV 3.0-rc1, and i have a postgres problem with too many postgres processes, eg : postmaster: navwrite manage 127.0.0.1 idle (postgres) postmaster: navwrite manage 127.0.0.1 idle (postgres) postmaster: navwrite navprofiles 127.0.0.1 idle (postgres) postmaster: navwrite manage 127.0.0.1 idle (postgres) postmaster: navwrite navprofiles 127.0.0.1 idle (postgres) postmaster: navwrite manage 127.0.0.1 idle (postgres) postmaster: navwrite navprofiles 127.0.0.1 idle (postgres) When i have more than 30-35 processes, then NAV stops working, and i have to stop and start the database. Looking into my postgres log file i see this : postgres[67232]: LOG: unexpected EOF on client connection postgres[67233]: LOG: unexpected EOF on client connection postgres[67234]: LOG: unexpected EOF on client connection postgres[67235]: LOG: unexpected EOF on client connection postgres[67236]: LOG: unexpected EOF on client connection postgres[67237]: LOG: unexpected EOF on client connection Anybody have a clue ? -- Med vennlig hilsen / Regards; Roar Pettersen Universitetet i Bergen - The University of Bergen Nygardsgt. 5 - N-5020 BERGEN - Norway Tlf: +47 55 58 40 55 VIP: 81503 fax: +47 55 58 40 70 roar.pettersen@it.uib.no - IT-Avd, UiB - http://www.uib.no

From kreide at gmail.com Fri Jun 10 13:28:37 2005 From: kreide at gmail.com (Kristian Eide) Date: Fri Jun 10 12:28:53 2005 Subject: [Nav-users] Postgres : unexpected EOF on client connection In-Reply-To: <Pine.GSO.4.61.0506101108230.4532@alf.uib.no> References: <Pine.GSO.4.61.0506101108230.4532@alf.uib.no> Message-ID: <200506101228.41534.kreide@gmail.com>

I've installed NAV 3.0-rc1, and i have a postgres problem with too many postgres processes, eg :

NAV does use quite a few database connections (on our production machine here I see 17 active now) as each subsystem needs one or more connections (heavily threaded subsystems may use more than one connection) and each http process will also require a connection.

When i have more than 30-35 processes, then NAV stops working, and i have to stop and start the database.

This seems like too many though, unless you have apache configured to spawn a high numer of processes.

postgres[67237]: LOG: unexpected EOF on client connection

Yes, we get these as well. The culprit seems to be alertengine, probably it is not closing its database collections properly (and this also seems to indicate it is opening a new database connection each time it polls the alertq, which is highly inefficient). Andreas, can you take a look at this? I also get a few of these in the log when starting alertengine: 2005-06-10 12:20:52 ERROR: syntax error at end of input at character 90 -- Kristian