0

I have a server that has 5 GB Ram and 3k MHz quad core CPU. It runs CENTOS 5 and usually worked ok. For a month or so I've been having problems with it, the load average raises above 4 and parts of it are starting to fail (like the mail server) and when I look at it it sais 35-65% CPU is wa, thou if I look at the momery I have more than 1 GB free, so it should not be working with swap. The only clue I have is the kblockd process (and usually kjournald also) that eats around 25% of my CPU. The problems with my server usually start around the same time with this processes. Usually ends up with me restarting the server, it doesn't seem to want to pass in it's own. This happens once a day. Any ideea of what might be happening?

Thanks, Virgil

Virgil Balibanu
  • 141
  • 1
  • 7

1 Answers1

1

During those spikes try to use iotop command and see what process(es) consume the disk. btrace is also a handy tool.

You might need to perform yum install iotop blktrace before you have those commands available.

Janne Pikkarainen
  • 31,852
  • 4
  • 58
  • 81
  • I will try it next time and write the results, thanks. – Virgil Balibanu Feb 01 '12 at 13:31
  • And try to interpret the results by yourself, too. :) If, say, Postfix consumes plenty of I/O according to `iotop`, then your system might be under spam attack. – Janne Pikkarainen Feb 01 '12 at 13:33
  • so what I found is this: postgres: stats collector process and kjournald. From this I guess that postgress is writing and reading a lot from the harddrive and kjournald is not a problem. Any ideea why this process would work so much with the harddrive and why it would not finish? – Virgil Balibanu Feb 02 '12 at 16:04