pct assigns the same core to all LXC

[MisterY]

Hi,
I have an R710 with 24 threads. But Proxmox (6.3.2) assigns mostly the same core and just gives a f*ck about the other 23.

----------------------------------------------------------------------
pct cpusets
100020:                  8      11 12          16
100110:                                                    20
100112:                            12
100121:                  8      11 12          16
100130:                            12
100132:                  8      11 12          16
100138:                            12
100154:            5 6 7        11       14       17    19 20
100156:  0   2     5     8      11 12          16             21
100211:                  8      11 12          16
10040:                  8      11 12          16
100913:                            12
100930:                  8      11 12          16
10101:                            12
10128:  0   2     5   7 8      11 12          16             21    23
10129:                            12          16
  1013:                            12
10131:                  8      11 12          16
10140:  0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
  1018:                            12
10180:                            12          16
10202:                  8      11 12          16
10212:                            12
  1024:                  8      11 12          16
  1025:                  8      11 12          16
  1034:                            12          16
  1052:                  8      11 12          16
   106:            5                     14       17       20
  1113:                            12
   119:                  8      11 12          16
   217:                            12

Here you can see that core 12 is assigned for 28 times, while core 1, 3, 4, 9, 10, 13, 15, 18, 22 and 23 is only assigned to one single LXC. I don't think that this is how it should be. If LXC 100138, 1113 and 217 are running at 100 %, they will block each other, while the other 23 cores are idling. How can I solve this?

 

[aaron]

This was a bug that is solved with pve-manager-6.3-3.

Check if you see it in the available updates

 

[MisterY]

Thx. Do I need to restart the host?

 

[aaron]

No, if you don't see the situation improve automatically a little bit after the update, you can try to restart the pvestatd daemon

systemctl restart pvestatd.service

If there is a new kernel available, you should reboot the node though.