Console becomes unresponsive, all else fine.

U

unixgirl

Greetings
I have run into a very odd problem. I have a FreeBSD 6.2-STABLE server that has been running fine for some time. Recently however, after a reboot, the console will become unresponsive after about 15 minutes. All other functions are fine. You can still ssh in. I have tried replacing the MB and CPU. It has even happened while in the middle of typing.

I would really appreciate it if someone could point me in the direction of things to check.

Thanks!

Nicole
 
Run top. Maybe there's something locking up. But usually FreeBSD remains usable under very high loads and you don't even notice unless some process is using 90% or something.

Something may be generating a lot of IRQs, so try vmstat -i too.
 
Any clues in /var/log/messages?

If you suspect hardware for some reason, likelier candidates (in that they're more likely to fail) would be disk drives, PSU, and RAM - in that order.
 
Sorry for the long delay.. I got pulled into several other projects and forgot to post a reply.

The problem it seems turned out to be some interaction between Squid (which was in-front of the server) and Lighthttpd and very quickly all the sockets were becoming filled with TIME_WAIT 65000 of them!

The solution I am using at the moment is to limit lighthttd to 512 connections via server.max-connections=256 in lighthttpd.conf. This seems to work nicely.

When I have more time I need to see who and why things suddenly stopped playing well together.

Thanks for everyone's suggestions!
 
unixgirl said:
Sorry for the long delay.. I got pulled into several other projects and forgot to post a reply.

The problem it seems turned out to be some interaction between Squid (which was in-front of the server) and Lighthttpd and very quickly all the sockets were becoming filled with TIME_WAIT 65000 of them!

The solution I am using at the moment is to limit lighthttd to 512 connections via server.max-connections=256 in lighthttpd.conf. This seems to work nicely.

When I have more time I need to see who and why things suddenly stopped playing well together.

Thanks for everyone's suggestions!
Click to expand...
Typo - I meant - Limit lighthttd to 256 connections via server.max-connections=256

Forgot to add.. that I find it sad/odd that having all the network sockets used up would cause the console to become unusable as well. The server was in no way out of memory and the console became unusable before sshing in did.
 
unixgirl said:
The problem it seems turned out to be some interaction between Squid (which was in-front of the server) and Lighthttpd and very quickly all the sockets were becoming filled with TIME_WAIT 65000 of them!
Click to expand...

Sounds like a Denial of Service attack, possibly caused by a virus infection...
 
You must log in or register to reply here.
Back
Top