Since I had no replication running,
it was enough to killall pvescheduler until no pvescheduler was running anymore,
Then service pvescheduler start.
And at least migrate and remograte the VMs, that the new libpbs was used.
A reboot was not needed in my case.
I removed the openvswitch stuff from one node and tried it again.
Unfortunately this doesn't solve the problem.
To be more precise:
If I move a windows server from an old node to an old node and back, it works.
If I move a windows server from an old node to a new node it works.
Some new tests:
The migration to cpu11 was working. The problem starts when migrating back to cpu02.
cpu02 uses openvswitch, while cpu11 didn't.
It looks like that the migration from a node without openvswitch to a node with openvswitch results in this behaviour.
Always when we migrate Windows Servers they stuck with an error screen.
After a reboot they work normal again.
Today I migrated some VMs to an other node and back to update the pbs library which is used by the VMs.
All 4 VMs needed a restart afterwards.
It was a DomainController, 2...
On friday I created a sync job to copy files from PBS-02 to PBS-01
PBS-02 has different datastores and PBS-01 uses the 'new' namespace feature.
Both PBS are running the latest no-subscription version 2.2.5
Updated 29.07.2022 and both rebootet,
On PBS-01 (target) I get...
You are lucky, I found a node where it still is the case:
root@pm-cpu-173:~# ps ax|grep pvesche
390618 ? Ss 61:21 pvescheduler
1271153 ? Ss 29:05 pvescheduler
2016484 ? Ss 50:23 pvescheduler
2711818 ? Ss 6:58 pvescheduler
3602118 ? Ss 43:26...
It's working now as expected.
kill -9 xxxx
was not needed.
terminates all processes.
I found more pve nodes where multiple instances of pvescheduler was 'running'.
How can this happen?
And why does a start of pvescheduler does not simply terminate all old running...
There is no tree, only single processes:
F UID PID PPID PRI NI VSZ RSS WCHAN STAT TTY TIME COMMAND
1 0 2916892 1 20 0 333232 19788 hrtime Ss ? 39:05 pvescheduler
4 0 608389 1 20 0 15504 5396 do_epo Ss ? 0:02...
journalctl (from the day where I restarted the pvescheduler):
Jul 21 18:04:31 pm-cpu-175 systemd: Stopping Proxmox VE scheduler...
Jul 21 18:04:32 pm-cpu-175 pvescheduler: received signal TERM
Jul 21 18:04:32 pm-cpu-175 pvescheduler: got shutdown request, signal running jobs...
I did this systemctl restart pvescheduler on friday.
Still the same message in the log of the backup job:
file /etc/pve/storage.cfg line 158 (section 'pm-pbs01-ednt') - unable to parse value of 'namespace': unexpected property 'namespace'
As written in #18, I did a
systemctl reload-or-restart pveproxy pvedaemon pvescheduler
It's getting worse. Now a few VMs more (also from other nodes) are not respect the namespace (which were Ok last times):
And I can still see:
file /etc/pve/storage.cfg line 158 (section 'pm-pbs01-ednt') -...