Question about only one vm on openvz

[b3rkl3y]

Hello,

First of all, i would like to thankyou for proxmox ve, it's great!

I'm making some tests with proxmox, I made a small cluster of 3 servers. 2 of them with multiple machines, and one of them with only one machine.

My main question is:
I know that openvz is designed to host multiple vm's, but tin my case, i would like to have a "dedicated server" for one vm. This would make my life easier, getting a good performance,easy backups, and the ease to move this vm from one host to another if needed.

Since in this server there will be only this vm, i would like to disable any kind of "brakes" or limitations (ok, only the ones that will increase performance) for this vm.
For example, in options I found the cpu units limited to 1000 what would be a good value for this? I increased to the maximum value (during heavy load), and i noticed that the physical host was very slow, even disappeared from the cluster now i changed to 50000 which would be a good option?

Messing around in the config files, i found an "unlimited template" (/etc/vz/conf/ve-unlimited.conf-sample) should I aply this to my vm? will my custom conf be kept when moving the vm from one server to another? or even when changing any option on the web interface? wich options should I change to unlimited?


And last question:
I found this on my logs, and i where wondering what was happening.

    proxwww    14658     Starting new child 14658
Jun 12 23:01:46     pvemirror    4334     starting cluster syncronization
Jun 12 23:01:46     pvemirror    4334     syncing templates
Jun 12 23:01:46     pvemirror    4334     cluster syncronization finished (0.15 seconds (files 0.00, config 0.00))
Jun 12 23:01:46     pvemirror    4334     restarting server after 2 cycles to reduce memory usage (free 84533248 bytes)
Jun 12 23:01:46     pvemirror    4334     server shutdown (restart)
Jun 12 23:01:46     pvemirror    4334     restarting server
Jun 12 23:01:52     proxwww    14695     Starting new child 14695

Those messages where repeating every minute, wich server was being restarted?

Thanks in advance!

 

[dietmar]

For example, in options I found the cpu units limited to 1000 what would be a good value for this? I increased to the maximum value (during heavy load), and i noticed that the physical host was very slow, even disappeared from the cluster now i changed to 50000 which would be a good option?

Just set it to 1000 (its a relative value). It makes no sense to set it to 50000 (dont do it).

Messing around in the config files, i found an "unlimited template" (/etc/vz/conf/ve-unlimited.conf-sample) should I aply this to my vm? will my custom conf be kept when moving the vm from one server to another? or even when changing any option on the web interface? wich options should I change to unlimited?

I would use the default settings instead.

Those messages where repeating every minute, wich server was being restarted?

The cluster sync server 'pvemirror' - you can safely ignore that message.

 

[fortechitsolutions]

Just a brief footnote,

If you want more detail on OpenVZ resource control, probably reading some online manuals at the OpenVZ wiki would be a good start (http://wiki.openvz.org/Main_Page)

Also to comment / agree with the prior post on this thread,

- the cpu units are relative, as a means to assign balance when multiple openvz VMs exist on one host. So a vhost assigned 2000 units gets 2x more cpu than a vhost assigned 1000 units. But the numbers used (1000, 2000, etc) - have no absolute or intrinsic value (ie, in terms of bogomips, actual mhz on the phyislal host, etc) - it is simply a 'counter' to partition the resource.

- also to comment (at least my understanding) - is that unlike PVM based vms, your openvz vhosts will see *all* cpus present in the physical host (so if a dual-quadcore host with 8 cores, then your vzhosts will see all 8 cpu cores...) (sometimes this detail is not immediately apparent when you first start using openvz VMs)

- additionally, resource allocations are more relevant when there is actual contention for resources. So if you have a physical host with 4 openvz VMs, and the total combined utilization never hits 100% CPU - then 'hard limits' are never imposed, because there is no need. The allocation limits only come into play when actually required (ie, VMs start demanding more CPU resources than are physically available at a given instant) - and then at such time, the openvz kernel starts assigning / restricting access to resources based on the resource allocation configs of the various VMs.

In your case, if you have only one VHost on the single physical host, then clearly there will never be resource contention, so it becomes less relevant. The only 'tunable parameters' that then become possibly of a concern might be caps in the OVZ Vhost config which might become limiting in the case of extreme loads within the vhost. Typically monitoring the 'beancounters' (see http://wiki.openvz.org/Proc/user_beancounters for more detail) will give you hints in case something is actively failing within a VM due to resource restraints that are imposed ...) But my guess is that the stock parameters used by ProxVE will be 'generous enough' that you should not have issues. (In the past I've seen issues when using 'very minimal' OpenVZ templates as stock for deployment; but this is NOT the stock config used by ProxVE for its openVZ host template..)

Anyhow. All this to ultimately say, that there is a lot of online documentation available at the OpenVZ website, and you might get more concrete details there if you think you need more info

Hope this helps slightly,


Tim

 

[b3rkl3y]

Thanks alot dietmar and fortechitsolutions

Both explanations where very clear and very complete.

I'm still impressed with the quality of proxmox ve. (of course that i'm also waiting for the new features on the roadmap).

Thanks