[SOLVED] Proxmox cluster slow to shutdown.

[ozdjh]

We're still forcing a "swapoff" before we reboot during an upgrade.

 

[Matthew Daniel]

We're still forcing a "swapoff" before we reboot during an upgrade.

We do the same before any reboot

 

[Bidi]

I have the same problem, 6.3-6

And even the strange think is even when i whant to start/stop a VM, in takes 1-3min and sometimes fails, this is strange, never before i update it to 6.3-6

trying to acquire lock...
TASK ERROR: can't lock file '/var/lock/qemu-server/lock-321.conf' - got timeout

 

[dragonslayr]

Curious if any of you were ever able to resolve this.
Running version 6.4-13

 

[Matthew Daniel]

PVE7 appeared to resolve it for me

 

[rogierl]

Well, for me it started to happen after upgrading to v7.x. Currently I'm running version 7.1-6. Is there a way to check what is causing the slow shutdown/restarts?

 

[t.lamprecht]

Hi.

Well, for me it started to happen after upgrading to v7.x.

This is already a thread with a lot of different stuff in, same symptoms but often not the same cause.

Is there a way to check what is causing the slow shutdown/restarts?

Can you please check the end of the journal of the last boot? journalctl -b-1 and then hitting shift + g to go all the way down and then scroll up from there. Would be interesting to see what units need that long to shut down, or if there's any dependency issue in systemd's unit configuration on your setup.

 

[rogierl]

Hi.


This is already a thread with a lot of different stuff in, same symptoms but often not the same cause.

Can you please check the end of the journal of the last boot? journalctl -b-1 and then hitting shift + g to go all the way down and then scroll up from there. Would be interesting to see what units need that long to shut down, or if there's any dependency issue in systemd's unit configuration on your setup.

See the logs here. I notice this:

Dec 02 08:40:07 pve systemd[1]: pvedaemon.service: Consumed 3min 31.155s CPU time.
Dec 02 08:41:18 pve pvescheduler[901]: server stopped

Now this isn't VERY long, I have to admit . But it feels like this took longer than before. So basically the question is: is this normal?

 

[t.lamprecht]

The consumed time systemd reports is itself not an actual issue, as that's rather the CPU time that the service used during its lifetime.

But there was actually a small regression in the stop behavior of the relatively new pvescheduler, it may wait a 75 seconds extra on its workers to finish even if they did already do so, that's fixed in pve-manager version 7.1-7 which should move to the enterprise repository today.

 

[rogierl]

The consumed time systemd reports is itself not an actual issue, as that's rather the CPU time that the service used during its lifetime.

But there was actually a small regression in the stop behavior of the relatively new pvescheduler, it may wait a 75 seconds extra on its workers to finish even if they did already do so, that's fixed in pve-manager version 7.1-7 which should move to the enterprise repository today.

Dude! Thanks man!! Proxmox rocks!
So happy with my subscription here... everybody should do that if running Proxmox on their home server.