Doubled load on Proxmox 4.1 compared to 3.4

W

WhoCares?

New Member
Jan 28, 2016
3
1
1
58
Short summary: On identical host machines the same VM workload produces a more than doubled load under Proxmox 4.1 than under Proxmox 3.4 before.

Details:
Hardware:
  • Dell M630 w/ 128 GB RAM, 2x Xeon E5-2650v2 @ 2.6GHz (Octacore)
  • Storage via NFS on a NetApp Metro Cluster
Proxmox 3.4 running 17 VMs:
node106:
top - 11:07:54 up 1 days, 16 min, 1 user, load average: 2,27, 2,15, 1,64

Proxmox 4.1 running the same 17 VMs:
node107:
top - 10:48:20 up 19:54, 1 user, load average: 5.20, 5.78, 5.96

There has been no change to the virtual disks since those are the same, migration has been done by copying over the *.conf files to the target node. Load within the VMs has been the same and the behavior is consistent and could be reproduced on other nodes with different numbers and workloads of VMs. The increase in the reported load of the host machine always is higher by a factor of 2.x on Proxmox 4.1 than it was on Proxmox 3.4.

Any idea what causes this behavior?
 
WhoCares? said:
Short summary: On identical host machines the same VM workload produces a more than doubled load under Proxmox 4.1 than under Proxmox 3.4 before.

Details:
Hardware:
  • Dell M630 w/ 128 GB RAM, 2x Xeon E5-2650v2 @ 2.6GHz (Octacore)
  • Storage via NFS on a NetApp Metro Cluster
Proxmox 3.4 running 17 VMs:
node106:
top - 11:07:54 up 1 days, 16 min, 1 user, load average: 2,27, 2,15, 1,64

Proxmox 4.1 running the same 17 VMs:
node107:
top - 10:48:20 up 19:54, 1 user, load average: 5.20, 5.78, 5.96

There has been no change to the virtual disks since those are the same, migration has been done by copying over the *.conf files to the target node. Load within the VMs has been the same and the behavior is consistent and could be reproduced on other nodes with different numbers and workloads of VMs. The increase in the reported load of the host machine always is higher by a factor of 2.x on Proxmox 4.1 than it was on Proxmox 3.4.

Any idea what causes this behavior?
Click to expand...

Recently I migrated from 3.4 to 4.0/4.1 - I did not encounter any remarkable difference in CPU load.

Is it possible to have a look at all the running processes and check which one consumes significantly more CPU resources respectively are there additional processes?
 
@Richard: with all the new cgroups tasks there of course were more processes running than on 3.4. As for specific processes that were taking up a lot of resources: No, nothing that immediately hit the eye. It's just that the load reported by the OS was a lot higher than usual and also that users reported the VMs to be laggy/slower. That's why I was asking for pointers.

@jkirker: No, we reverted back to 3.4
 
You must log in or register to reply here.
Top Bottom