Portal Home > Knowledgebase > Articles Database > Server load peaking fixed by httpd restart??


Server load peaking fixed by httpd restart??




Posted by divine_lighting, 11-09-2011, 12:03 AM
Happens a few times a month, and twice within the last day. Web sites stop responding, I get a notice from DirectAdmin that server load is high. The problem usually "resolves" on it's own after a few minutes, though sometimes it does not. Restarting apache fixes it instantly. What/who is causing this? Here is most recent episode: Kernel 2.6.18-274.3.1.el5 Centos 5.7 Apache 2.2.17

Posted by fabin, 11-09-2011, 12:09 AM
May be, few sites are having high traffic. You have to check the access logs or use server status module ( mod_status ) in order to find which websites are having the most accesses.

Posted by SolidJoe, 11-09-2011, 12:20 AM
That is a lot of mail and mysql processes. I'd take a look there.

Posted by divine_lighting, 11-09-2011, 01:00 AM
I own all the websites on the server and there is no known substantial traffic to them, neither should there be much email action. There is no outgoing email this time of day, and maybe a few incoming emails per hour. Gmail checks two accounts. I'm a newbie. How should I investigate dovecot and mysql activity? Thanks.

Posted by divine_lighting, 11-09-2011, 07:10 PM
Just happened again. Suggestions welcomed.

Posted by fshagan, 11-09-2011, 09:38 PM
Do you own an iPhone, iPad, Mac or other device using IMAP email services? Set the preferences to check email every 5 minutes rather than the default of every 2 seconds or whatever it is. Apple + IMAP will seriously impact your mail server. You can also install CFS from ConfigServer.com to block IPs after 5 attempts at logging in; I can't tell from the logs, but you might be getting hammered by brute force intrusion attacks.

Posted by ssfred, 11-10-2011, 09:38 AM
Check whether the server is being attacked by using the following command netstat -ntu | awk '{print $5}' | cut -d: -f1 | sort | uniq -c | sort -n This will provide the no. of connections from each IP. Also install an effective mod_sec rule set along with CSF and LFD as mentioned in another post.

Posted by ArturasLIX, 11-10-2011, 10:31 AM
Hi, output that you provide does not help much. As I can see from vmstat command, at that time your system was idle. Where is no running processes. So at the time you ran your commands your system was completely idle. So it's difficult to determinate problem with such info. I suggest you to run TOP command when load rises and paste output here, so we can see ram usage, running processes count. If there is few web sites on server, try to run tail -f /path/to/logs/*.log to see web sites activity. Perhaps you will find one or more IP addresses to block. And could you provide exact message from DA?



Was this answer helpful?

Add to Favourites Add to Favourites    Print this Article Print this Article

Also Read
Weird DNS problem (Views: 672)


Language: