swap memory 100% after upgrade to 9.3

[Miguel Mesa Mesa]

My server:
CentOS 5.4 x86_64 4GB RAM Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz

After upgrade Plesk from 9.2.3 to 9.3, in 24 hours the RAM was at 99% and SWAP (4 GB) was at 100%. So I rebooted the system and after RAM is 25% and SWAP 0% used.

Some hours after the used RAM is growing slowly, now it is in 50%. I think I have to reboot again tomorrow.

I search for process eating RAM or CPU, but I don't find it.

 

[lvalics]

This is why usually is good to wait 2-3 days to see if no problem popup

 

[atomicturtle]

Memory leak huh? Run top and hit M (this sorts by memory usage)

 

[Miguel Mesa Mesa]

What do you think about this?

 

[Miguel Mesa Mesa]

I had to reboot the server again, the resources are:
3% CPU, 75% RAM, 100% SWAP

The image files of top command are after reboot the system.

 

[atomicturtle]

Are you using a non-standard kernel by any chance? Ive seed that sw-engine-cgi runaway before with them.

 

[Brujo]

this amount of processes for sw-engine-cgi is not normal. Afaik there was a issue in previous Plesk versions. Try to find the solution here in the Forum...

Brujo

 

[Miguel Mesa Mesa]

Thank you, atomicturttle and Brujo.

I searched in the forum, and I find a similar case in thread http://forum.parallels.com/showthread.php?t=85436 (OVH CentOS release, 100% SWAP...), but there is not a solution...

I have the OVH CentOS 5.4 release, that perhaps has an incompatibility with kernels with grsec (I don't know what is it), like it's described in above thread. But, I don't Know how to indentify if my server has that problem and how to fix it.

I executed:
killall -9 /usr/bin/sw-engine-cgi
/etc/init.d/psa start

and RAM and SWAP used go down to 50%.

Can anybody help me?
(excuse me for my english)
Thanks in advance.

 

[atomicturtle]

Thats correct, its a problem with the OVH kernel design. Its something we fixed with the ASL kernels when 9.0 first came out, unfortunately its not something you can fix with the OVH kernels without rebuilding them from scratch.

 

[Miguel Mesa Mesa]

If I understand it, changing the OVH Kernel for a CentOS repository kernel maybe fix the problem?

What is the correct yum command? And, which are the caused problems by changing kernel?

 

[atomicturtle]

Unless OVH started creating the correct modprobe.conf files, then no its not going to drop right in. You'll need to determine the appropriate modules for your running kernel. We have a few wiki entries on the topic in the atomicorp wiki, but they are by no means comprehensive:

http://www.atomicorp.com/wiki/index.php/Kernel
http://www.atomicorp.com/wiki/index.php/Kernel_Panic

 

[Miguel Mesa Mesa]

Thank you very much.

I found a very easy solution: use the netboot service of OVH Manager. I can choose the best kernel without change any file.