0

I am running a lot of openvz container in server. Now I have a lot of performance issue, and I think they come from I/O since munin graph show a lot of I/O activiy.

Server is proxmox, and it was recently upgraded (this could be the problem also) But near upgrade to last version of proxmox I added a gateway in a container running shorewall.

But now is almost impossible to log in the server, rebooting makes the system usuable but after two or tree hours I/O problems come again.

There isnt too much network activiy (10kb/s) it's a home "test" server.

Any other idea in order to detect the problem? openvz container asgateway is not a good idea?

Improve this question

2 Answers 2

Reset to default
0

You may look at the "failctl" column by typing this command :

vzctl exec 101 cat /proc/user_beancounters

You now need to adapt (101 here) this number to all of your VMs. If some beancounters are not equal to 0, you need to increase those value.

Now for example, if you want to grow a value on the fly, you could do like this :

vzctl set 101 --privvmpages 786432:1048576 --save --setmod restart

Now, for your I/O poblems, on the main machine (hosting VMs), is a "top" commands is giving you some iowaits ? If it's the case, it means you should add more disks.

Improve this answer
4
  • yes, at hardware node there was a problem with a process using %100 cpu. I now change the kernel, but not sure if this would fix the problem since is like random
    – llazzaro
    Feb 19, 2010 at 13:36
  • Did you try to upgrade to the last version of proxmox ?
    – Deimosfr
    Feb 19, 2010 at 20:23
  • the problem started after upgrade to last version (1.5)
    – llazzaro
    Feb 21, 2010 at 23:12
  • Did you look at beancounters like I told you ?
    – Deimosfr
    Feb 22, 2010 at 9:58
0

I switch back to kernel 2.6.18 and the I/O problem went away. It seems a problem with kernel 2.6.24

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .