What should I look for after server down

[baako]

Hi,

Today, my server were down, even I performed a hard reset, the web site is still not working. Everything solve when I use the reboot command from the shell.

May I ask will there any logs I should look for to find out why the server were down suddently? Anything I can do to avoid this happen again?

Thanks. Oh! btw, I am using FreeBSD and Plesk 7.5.2. Thanks.

Regs,
Baako.

 

[editor]

I would look at first into

# ls -l /usr/local/psa/admin/logs
# vi -R httpsd_error_log
# vi -R httpsd_access_log

 

[phingers]

same problem here

The server over the past week reboots every morning, for some unknown reason, and today it rebooted at 11 and hung at 1230ish

I pressed the reset button on the server at 12:30

It comes back up fine, there is nothing in the log files, my guess its either qmail, mysql, or apache.

Nothing in the log files in /var/log or the admin ones referrered above.

Ideas and next steps to troubleshoot?

Perhaps its a vhost log file that may shed some light, but there are over 50 domains on my server. How do I isolate it, or restrict apache processes to 80% cpu or something?

 

[Linulex]

sounds to me a lot like a hardware problem, most likely memory.

you could try this in the httpd.conf:

RLimitCPU 8 12
RLimitMEM 15000000
RLimitNPROC 10


Place it in the Main server configuration section. These limits work for us, you should experiment to find your own best limits.

regards
Jan

 

[phingers]

hopefully I figured it out. memory

I had the colo replace the RAM yesterday, and the box has been up 22 hours now.

The memory free is down to around 13000k which is quite low, and it hasnt crashed or hung.

From what I understand redhat automatically uses up the memory, and dishes it out when apps need it, so the amount free is misleading.

In the future though are there any mem tests or something I can do on the hardware, like we can do in windows?

I appreciate everyones help, thanks again for all the suggestions.

 

[Linulex]

The actual free ram is
mem used - (mem cashed + mem buffered)

you can see the complete layout of mem by doing this at the console (or in ssh):

cat /proc/meminfo

We use memtest86+ to test all the memory before it goes life.

http://www.memtest.org

regards
Jan