vzcpucheck & sparc problem [message #16311] |
Fri, 31 August 2007 11:02 ![Go to next message Go to next message](/theme/ovz3/images/down.png) |
blam
Messages: 1 Registered: August 2007
|
Junior Member |
|
|
I'm using OpenVZ 2.6.18-028stab039 (Ubuntu OS) on a Sun Fire T2000 machine.
Everything else but vzcpucheck seems to work just fine. Vzcpucheck thinks the processing power of the hw node is 0 and I believe this is the reason why I'm getting really poor performance out of my VE's. Here's the output of vzcpucheck-command:
root@hostname:# vzcpucheck
Current CPU utilization: 5000
Power of the node: 0
Warning: hardware node is overcommited
I have 4 VE's running with each of them having default cpuunits (1000). There's zero amount of extra load.
Is this some kind of bug or what?
[Updated on: Fri, 31 August 2007 11:03] Report message to a moderator
|
|
|
Re: vzcpucheck & sparc problem [message #16312 is a reply to message #16311] |
Fri, 31 August 2007 11:58 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) |
|
It's just that vzcpucheck is not able to parse SPARC's /proc/cpuinfo. Not a problem at all, since cpuunits are values relative to each other.
In other words, a VE gets an amount of CPU which is equal to it's cpuunits divided by sum of all VE's (including VE0) cpuunits.
So, in your case you have 4 VEs with cpuunits=1000, plus VE0's cpuunits of 1000, that means each VE will have 1000/5000 = 0.2 = 20% of CPU time (in extreme case when all VEs will be busy). That means 20% is a guarantee, and there is no limit (unless you explicitly set it with --cpulimit).
So, again, your setup is fine, and each VE can get anything from 20% to 100% of CPU time, depending on the load.
Kir Kolyshkin
|
|
|