crazy server load related to eximstats?

[SMG]

I have a problem with one of my servers where randomly (could be 0-3 times a day) the load goes completely out of control (I once saw it hit 95). I look at process list and see a lot of mysql's, and figure ok I could restart that maybe to knock off some usage, but mysql won't restart AT ALL if I do that at that time (no method of restart works unless I do a hard reboot).

When it's like that, also rebooting via whm, ssh or master reboot switch just makes the server hang, it doesnt actually reboot - I have to have a DC tech manually reboot it.

For some reason the first time I saw it I suspected eximstats (I dont know why, but I do remember seeing its LOCK TABLES command sitting in mysqladmin processlist), however I also noticed ... last night the same exact thing happened. I did /scripts/restartsrv_eximstats and the load instantly began dropping from around 30 to about .25-.5 ... so obviously it has something to do with eximstats (I also tested with a few others like restarting apache and exim, but that didnt reallly help). I have a few ideas about this but before messing with my server any more I'm curious if anyone else had and fixed this problem?

[cpanelben]

Eximstats solely shouldn't be the cause of your load issues. One thing that I would definately check is file system corruption. If you haven't already fsck'd the drive(s) on the server (and it's a Linux OS), then I would definately recommend that. It'll require a reboot, but it shouldn't take more than a few minutes of downtime. If it's a redhat based OS, then you can accomplish this by touching the file /forcefsck and rebooting the server. You can also accomplish the same thing by giving shutdown the "-F" flag, such that:

shutdown -rF hh:mm

See "man shutdown" for more information on this command.

If the fsck doesn't help, then have your DC or NOC check the system for the cause. If it comes down to a software issue, then open a support ticket with your license provider.

[jcase]

yup, I too can tell you that exim is causing a considerable load on my box. How can one renice all the processes so that it DOESNT run as priority -6? I dont care about email nor do my clients as much as they care about their sites being up and loading quickly. The priority of sending email doesnt concern me.

[cpanelben]

Eximstats and Exim are not the same process. Please create a new thread concerning Exim process priority. And generally, Exim doesn't use much processesing power sending messages, it's load level is mostly determined by the number of messages it receives.

[SMG]

Eximstats solely shouldn't be the cause of your load issues. One thing that I would definately check is file system corruption. If you haven't already fsck'd the drive(s) on the server (and it's a Linux OS), then I would definately recommend that. It'll require a reboot, but it shouldn't take more than a few minutes of downtime. If it's a redhat based OS, then you can accomplish this by touching the file /forcefsck and rebooting the server. You can also accomplish the same thing by giving shutdown the "-F" flag, such that:

shutdown -rF hh:mm

See "man shutdown" for more information on this command.

If the fsck doesn't help, then have your DC or NOC check the system for the cause. If it comes down to a software issue, then open a support ticket with your license provider.

As far as I know, it has been fscked recently, because I did have a bad drive before but that drive has been unmounted and I havent had other problems since then, and when the bad drive was there techs manually rebooted and fscked both drives. The thing I don't understand is ... whereas normally I would think it could be a hard drive problem or something, this ONLY happens maybe once or twice a day tops (it hasnt happened in 1.5 days now) and EVERY time its the same thing, mysql goes out of control, and restarting eximstats fixes it ... if it was an hd problem wouldnt more than eximstats be to blame? I run batch scripts and I know mysql works fine under pressure from them, so I dunno if its anything related to the sectors on mysql, however I also noticed most of the mysql processes during this time were opening or closing tables...

[cpanelben]

You should also check the database for errors. Look at the myisamchk command for more information. Here's MySQL's doc on it:

http://dev.mysql.com/doc/mysql/en/repair.html

A google search for "mysql repair" will give you all sorts of good info.