Hi Chris,
thanks for your help.
Chris Elsworth wrote:
> On Wed, Dec 28, 2005 at 09:42:56PM +0100, Dennis Thomas wrote:
>>on a machine acting as a loadbalancer for some hosts the pen proccess
>>dies every 1-3 Minutes. The last thing i can get in the log is "Error
>>opening socket".
> EMFILE: potentially a problem, check with ulimit -n for this. It's
> probably in the order of a few thousand, so unlikely.
...
> ENOBUFS: most likely the cause; try netstat -m and see if your mbuf
> clusters are maxed out next time it happens. Raising maxusers will
> allocate more mbufs and thus fix this. (kern.maxusers)
i changed the source to get the error code, it's 55 (whatever this may
mean...).
I checked the mbuf clusters, they are not the problem:
6043/25600 mbuf clusters in use (current/max)
I'll compile a new kernel with fewer open files and see it it helps.
Thanks,
Dennis
Received on Wed Dec 28 2005 - 23:09:05 CET
This archive was generated by hypermail 2.2.0 : Wed Dec 28 2005 - 23:09:06 CET