Hi,

My server has strange behaviour and I don't know if this is by design or if 
there is something wrong. 

It has around 60.000 pop accounts, load average is around 12-13, but 
response time is around 2 seconds. I run Linux, lastest Redhat kernel 
2.2.19-6.2.1 and Bruce's rpm + daemontools + supervise-scripts-3.3. 

I'm monitoring the behaviour of tcpserver for the pop service ( simple 'ps -
aux | grep tcpserver | grep pop' and I get, obviously, just one is 
running ). Sometimes, supervise tries to restart tcpserver and I see 2 
tcpservers. Then the second dies ( port is already bind ). During this 
time, pop fails. Looking at the log for the pop service, I can tell that 
there is no specific moment that it happens. Tcpserver didn't run out of 
connections ( I have up to 150 concurrent ), sometimes it did with 20 
connections, sometimes with 90 concurrents. 

Is this an expected behauvior ? Is there a way to tune up supervise ? ( 
time-out, any other parameters ? )

Thanks in advance
Renato - Brazil

Reply via email to