Very large IO wait percent

[ffeingol]

First off, we
e not sure that this is related to CL at all, but we are just trying to eliminate different factors.

We have a server running cPanel + CL + LSWS. The IO wait percent is extremely high. There are a ton of busy forums on this server. We tried switching from Apache to LSWS and that has leveled out the server load, but the IO wait is still very high. From what we can track down its MySQL that is causing the heavy IO.

We have similar servers (same hardware) without CL and they are not having the issue. We are running 2.6.18-374.12.1.el5.lve0.8.54PAE (32 bit). My questions:

1) Are there any know issues with CL that would cause this high IO wait?

2) Is it safe / possible to reboot back to a stanard CentOS kernel (at least for a while) to rule out CL? Would we have to re-run easyapache if we do this?

Thanks

[iseletsk]

We haven found evidence that standard CentOS kernel gives better IO performance, but it is totally safe to reboot to standard CentOS kernel.

[iseletsk]

Also, try switching to deadline IO scheduler. It might help.

[info]

Old thread I know... but switching to deadline IO scheduler (instead of cfq) greatly improved page load times on our server!