up vote 1 down vote favorite
share [g+] share [fb]

My server is a FreeBSD system, I run nginx on it to proxy user requests to another real server. My problem is I can't visit my freebsd server after a while. The server is in another place, so I have to go to fix it. When the problem happens, the server can't ping another address, it can't use networks, if I reboot it and restart nginx, everything becomes normal. But after a while, it can't be visited again. I am not quite familiar with freebsd, so anyone can't help to find out the possible reason?

link|improve this question
feedback

3 Answers

up vote 1 down vote

I think it's problem with mbufs count. There is 2 possible solutions:

You can easily check limits: Compare values LIMIT and USED in vmstat 

# vmstat -z
link|improve this answer
Do you mean the server has used out of all mbufs? I guess this will happen if we've too many visits, but our server is not visited by many users(not more than 100 before it's down), and will mbufs be reused after a time, which means the server will recover after some time? Till now nothing changed actually. – gleery Oct 20 '09 at 8:59
mbufs, open files, somaxcon etc... there is too many limits in OS that you can bump in. – SaveTheRbtz Oct 20 '09 at 9:04
thanks for your help, I will check by your hints – gleery Oct 20 '09 at 9:11
feedback
up vote 0 down vote

Check the network card as well - it might be the mbufs issue that SaveTheRbtz mentions but I've also had this problem with a network card that would inexplicably 'wedge' after a while.

link|improve this answer
Yeah, that's might be good suggestion. Grep /var/log/messages for something like watchdog time out and other messages from your NICs' drivers – SaveTheRbtz Oct 21 '09 at 20:58
feedback
up vote 0 down vote

Also check out your firewall, I had some problems with a server not forgetting about connections for a few hours. Eventually it would fill up its allotted connections and just reject everything else. If I waited it would come back. In the end I just had to disable connection tracking.

link|improve this answer
pf have state table that can overflow: use pfctl -sa and watch LIMITS – SaveTheRbtz Oct 20 '09 at 20:27
feedback

Your Answer

 
or
required, but never shown
discard