0

I've got a CentOS based Plesk VPS which keeps on slowing to a crawl.

I checked what was happening with top, and found my load averages seem to be frequently spiking to very high numbers.

After watching top for some time, when my CPU usage spikes to 100% or close to it, I can only count around 10% of CPU usage in the process list, and can't see what is using up the remaining 90%.

I've captured some output from top using

top -b -i -n 20 >> ./top_procs


top - 09:21:46 up 19:09,  5 users,  load average: 6.42, 4.02, 3.39
Tasks:  55 total,   2 running,  52 sleeping,   1 stopped,   0 zombie
Cpu(s): 81.2%us, 18.8%sy,  0.0%ni,  0.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Mem:    524288k total,   360524k used,   163764k free,        0k buffers
Swap:        0k total,        0k used,        0k free,        0k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                                                                
22227 apache    19   0  111m  38m  14m R  4.7  7.5   0:00.19 httpd                                                                                                  
21721 root      15   0  2140 1028  808 R  0.0  0.2   0:00.00 top

I'm not totally sure I'm looking for the right thing here, but this is representitive of what is happening in general.

Any thoughts on how I can locate the cause of the slow down would be greatly appreciated! I've searched Google thoroughly to no avail!

Thanks,

Ian

soopadoubled
  • 183
  • 2
  • 8

1 Answers1

1

In top, press "1". You may find a single processor is overloaded.

Normally, I see this sort of thing with iowait, but yours is at zero. Still, 

iotop

may be informative.

BMDan
  • 7,249
  • 2
  • 23
  • 34
  • My VPS has been allocated a single processor. iotop is not installed my system, so I'll install it and give it a try. Thanks. – soopadoubled Jul 22 '10 at 11:36
  • I can't put iotop on my CentOS 5.x box, but I have used dstat and iostat, and everything on the i/o side looks fine. I've tried running ps -aux when top shows my CPU spike, but the processes only add up to around 10% of CPU usage. I'm on a single processor VPS, so there's no twisting of the figures from running multiple processors. Any further thoughts would be greatly appreciated! – soopadoubled Jul 28 '10 at 15:58
  • Load also includes time spent in uninterruptable sleep. One common source is NFS mounts. Got any of those? Separately, you never really mentioned--is the box responsive during all of this? – BMDan Jul 29 '10 at 16:09
  • I've not got any of those. Further more the box becomes unresponsive during all of this. Even SSH slows to a crawl. Thanks for your input - but I moved to a new server and started afresh, without Plesk eating up resources. – soopadoubled Aug 12 '10 at 10:06