Quantcast
Channel: VMware Communities : Popular Discussions - VMware Server Archives
Viewing all articles
Browse latest Browse all 69891

tracing iowait

$
0
0

I have some question about a performance problems:

 

we set up a vmware server on top of a SuSE 10.1 system

with two opteron CPUs and 8GB of memory. Running vmware

server, we currently run about 16 guest VMs, mainly running

different kinds of Windows systems. Most of them are more

ore less idle. Frequently, however, the system starts

thrashing with 20-50% iowait, and users of the different

vmware clients complain about "system unusable". If this

continues, I have to stop my VMware experiments, before

people are killing me...

 

The problem is: i can't find who causes this excessive

iowait. I see 4 cpus almost idle, there is about 6GB

memory available. "top" shows a few vmware-vx processes

at 5-8% CPU. Using "iostat" i see a high IO rate on the

partition containing my VMware images, but I can't assign

this IO to any process.

 

All vmware images reside on an ext3 system on an 3ware raid

controller with 6x500GB SATA disks. "hdparm -t" reaches about

270 Mb/s, so this is not really "slow".

 

Any idea how to find the bottleneck?

 

Dieter.


Viewing all articles
Browse latest Browse all 69891

Trending Articles