On Aug 23, 2006, at 6:26 AM, Jesse Lee wrote:
The system dead everyday, There are lots of Disk I/O oprations, We
can't login with SSH and Console when it dead. But when we reset the
system, Everything is back to normal. Can't find where is wrong.
We run slashcode on centos 4.3 with 2.6.9-34.0.2.EL kernel,and mysql
4.1.20, for redhat-linux-gnu (i686) using readline 4.3. and with perl
v5.8.6 built for i686-linux. Slashcode version is Slash T_2_5_0_112.
The hardware run in a vmware ESX 2.5 guest system. with 1G memory and
Check your processes - see if there's anything system wide that's hogging everything.
Check your swap space when "it dies", do you have enough? what's swapping?
Use vmstat and see if what's clogging is I/O or CPU that's the bottleneck.
Check your disk space, got enough free?
Check your httpd error log files.
Check your /var/log/*
Check your mysql error log.
Under mysql, do a SHOW PROCESSLIST to see what queries are hanging (if any).
I would recommend you use mysql5, first off.
You can enable your slow query log on mysql.
Read through some of those links, check your MaxClients and MaxSpareServers, compared to the amount of ram you have.
Hardware wise, you should be fine for a not-huge site. That's not far from our config here (except we always run mysql on it's own machine(s)). Since you seem to be running apache and mysql on the same machine, you should make sure that your mysql and your apache are setup such that you're not running out of ram.
For slash, there's options such as cache enable (see system vars) as well as on httpd startup preloading of things (look at your *site's* httpd.conf file). That can eat ram. And slashd can be a resource/ram pig when it runs, just because it does quite a lot. Look at it's processor and ram usage when it's running.
Be sure if you figure it out, to email the list and give details of what you found and how you fixed it so the next person who runs into such a situation can find out what you did via the listserve archives.