Don't limit vcpus for netvms to 1

[marmarek]

Reported by joanna on 16 May 2012 13:19 UTC
I don't see any reason why all netvms got vcpus set to 1...

Similiarly

Migrated-From: https://wiki.qubes-os.org/ticket/571

[marmarek]

Comment by marmarek on 18 May 2012 20:18 UTC
If you want >1 vcpus, just set it. It caused problems with some devices (doesn't remember details), so leaving one by default is reasonable choice. Anyway >1 vcpus in netvm/proxyvm will not be used in any normal use case.

[marmarek]

Comment by marmarek on 18 May 2012 20:18 UTC
If you want >1 vcpus, just set it. It caused problems with some devices (doesn't remember details), so leaving one by default is reasonable choice. Anyway >1 vcpus in netvm/proxyvm will not be used in any normal use case.

[marmarek]

Comment by joanna on 21 May 2012 08:22 UTC
When yo use local Gbps networking, what is CPU load in your netvm and firewallvm?

[marmarek]

Comment by joanna on 21 May 2012 08:22 UTC
When yo use local Gbps networking, what is CPU load in your netvm and firewallvm?

[marmarek]

Comment by marmarek on 21 May 2012 12:05 UTC
In netvm about 80% (top process netback/0), in firewallvm - 60%. This was single tcp stream of ~600Mbps (iperf).

[marmarek]

Comment by marmarek on 21 May 2012 12:05 UTC
In netvm about 80% (top process netback/0), in firewallvm - 60%. This was single tcp stream of ~600Mbps (iperf).

[marmarek]

Comment by joanna on 22 May 2012 09:17 UTC
80% is a lot... And how does it change when you enable 4 (all) VCPUs?

I really think we should not superficially disable all the VPCUs "just in case". Rather, all VCPUs should be assigned to all VMs by default, and only in case of some troubles, one could try to limit the number of VCPUs...

[marmarek]

Comment by joanna on 22 May 2012 09:17 UTC
80% is a lot... And how does it change when you enable 4 (all) VCPUs?

I really think we should not superficially disable all the VPCUs "just in case". Rather, all VCPUs should be assigned to all VMs by default, and only in case of some troubles, one could try to limit the number of VCPUs...

[marmarek]

Comment by marmarek on 25 May 2012 13:04 UTC
With assigned more vcpus (2 on my system), there is still only one netback process, and still it uses only one vCPU.
But ok, it is still possible to change number of vcpus assigned (in case of problems), so default can be "all". Maybe the problems (whatever it was) are now resolved in newer xen/kernel.

[marmarek]

Comment by marmarek on 25 May 2012 13:04 UTC
With assigned more vcpus (2 on my system), there is still only one netback process, and still it uses only one vCPU.
But ok, it is still possible to change number of vcpus assigned (in case of problems), so default can be "all". Maybe the problems (whatever it was) are now resolved in newer xen/kernel.

[marmarek]

Comment by marmarek on 31 May 2012 01:23 UTC
http://git.qubes-os.org/gitweb/?p=marmarek/core.git;a=commit;h=812264cfb042d47ce13059e3148522793e1cdd50

[marmarek]

Comment by marmarek on 31 May 2012 01:23 UTC
http://git.qubes-os.org/gitweb/?p=marmarek/core.git;a=commit;h=812264cfb042d47ce13059e3148522793e1cdd50