Interesting to notice I restarted james yesterday and put connections log on debug my server never receives more than 50 simultaneous connections and the connection log shows more connections been added to the pool
Today the log shows the following 18/02/04 12:04:43 DEBUG connections: Starting connection on 200.229.128.62:25 from 200.229.128.24:35892 18/02/04 12:04:43 DEBUG connections: Ending connection on 200.229.128.62:25 from 200.229.128.24:35892 18/02/04 12:04:43 DEBUG connections: Releasing one connection, leaving a total of 92 18/02/04 12:04:43 DEBUG connections: Returning a org.apache.james.util.connection.ServerConnection$ClientConnectionRunner to the pool 18/02/04 12:04:44 DEBUG connections: Ending connection on 200.229.128.62:25 from 200.162.208.35:3334 18/02/04 12:04:44 DEBUG connections: Releasing one connection, leaving a total of 91 As we can see the server is releasing some connections but some are not end the pool is growing little by little until it gets the limit in my case 150 connections. I will do a test i our developement server to see if I can reproduce this. []s On Wed, 2004-02-18 at 08:47, Michael Nestler wrote: > Danny Angus wrote: > > >>>>believe). As a strange aside the default*.log contains a database > >>>>connection exception (it contained the same exception at the last time > >>>>James stopped 6 days ago). I don't know if this means anything - could > >>>>this exception have caused the problem? spoolmanager*.log and > >>>>mailet*.log stopped recording at 1:32 am. > >>>> > >>>> > > > >I wonder if James is really failing because it is loosing db connectivity > >and therfore whilst it can open smtp connections it somehow can't complete > >the whole transaction quickly enough, connection pool grows to its limit > >and James chokes. The first obvious symptom would be the socket limit being > >reached, but the cause could be loctaed elsewhere. > > > > > > I would like to add that "cat /proc/sys/fs/file-nr" returns in my case > "3505 2878 209714" - so more than 200,000 sockets should be available. > Also, there is another web application running on the same server, using > the same database server and Internet connection. It never stopped > working when James did. > > If James would actually use 120 socket handles - wouldn't they appear in > the netstat output? I would say so - but perhaps I am wrong on this. I > believe James is only using 120 socket objects (Java objects) - but not > 120 Linux sockets (otherwise they would appear in the netstat output). > James (or the underlying pool implementation) doesn't seem to realize > that the socket is already closed. > > Michael > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] >
