| Looking for reason for high load [message #45481] |
Mon, 12 March 2012 19:48 |
mustardman
Messages: 91
Registered: October 2009
|
Member |
|
|
Been chasing this problem for awhile. Once in awhile I get high load. Usually it's easy enough to track down. I look at top and see if it's CPU or I/O related. Usually it's IO. I find the busy process using either top or iotop then find out which container it is using vzpid.
But this one has me stumped. I get high load number in top but I/O (%wa) is hovering around zero and CPU (%us) is also idling. Don't see any excessive process activity. Looking through logs in each container I don't see any ssh scanning or http scanning on any of the containers. Nothing unusual in the logs in general.
So no evidence of high load anywhere other than a high load number in top. One theory I have is because I have low I/O (2) priority in one container (lots of clamAV and spamassassin activity) that one container generates an artificially high load number when there is a lot of spam email activity because of the IO priority setting even though it's not loading down the system.
I remember reading somewhere that someone mentioned that this can happen if you try change I/O priority.
The kernel on this node is:
2.6.18-274.7.1.el5.028stab095.1 x86_64
Anyone else have any ideas how I can try trace this and if IO priority gives artifically high load numbers on CentOS 5 nodes?
I have not had a chance to try change the IO priority to see if it fixes this yet because these high load spikes are somewhat random and don't last too long. So I haven't had a chance to test this theory out yet.
I have also looked at the load reading inside the suspect container as well as all others and they all are very low. Not sure what else I should be looking for to find the source of the high load.
[Updated on: Mon, 12 March 2012 19:55] Report message to a moderator |
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
