[ClusterLabs] Antw: Re: Antw: [EXT] VirtualDomain ‑ started but... not really
Ulrich.Windl at rz.uni-regensburg.de
Tue Dec 14 10:05:04 EST 2021
>>> lejeczek via Users <users at clusterlabs.org> schrieb am 14.12.2021 um 15:48
Nachricht <653e9087-e648-9b8d-01de-341d2350f825 at yahoo.co.uk>:
>> My guess is that you checked the corresponding logs already; why not show
> them here?
>> I can imagine that the VMs die rather early after start.
>>>>> lejeczek via Users <users at clusterlabs.org> schrieb am 10.12.2021 um
>> Nachricht <df8eac8f‑a58e‑28e5‑53b5‑73eb1fe432b2 at yahoo.co.uk>:
>>> Hi guys.
>>> I quite often.. well, to frequently in my mind, see a VM
>>> which cluster says:
>>> ‑> $ pcs resource status | grep ‑v disabled
>>> * c8kubermaster2 (ocf::heartbeat:VirtualDomain):
>>> Started dzien
>>> but that is false, also cluster itself confirms it:
>>> ‑> $ pcs resource debug‑monitor c8kubermaster2
>>> crm_resource: Error performing operation: Not running
>>> Operation force‑check for c8kubermaster2
>>> (ocf:heartbeat:VirtualDomain) returned: 'not running' (7)
>>> What is the issue here, might be & how best to troubleshoot it?
>>> ‑> $ pcs resource config c8kubermaster2
>>> Resource: c8kubermaster2 (class=ocf provider=heartbeat
>>> hypervisor=qemu:///system migration_transport=ssh
>>> Meta Attrs: allow‑migrate=true failure‑timeout=30s
>>> Operations: migrate_from interval=0s timeout=180s
>>> migrate_to interval=0s timeout=180s
>>> monitor interval=30s
>>> start interval=0s timeout=90s
>>> stop interval=0s timeout=90s
>>> many thanks, L.
>>> Manage your subscription:
>>> ClusterLabs home: https://www.clusterlabs.org/
> Not much there in the logs I could see (which is probably
> why cluster decides the resource is okey)
> What is the resource's monitor for it not for that exactly ‑
> to check the state of resource ‑ whether it dies early or
> late should not matter.
> What suffices in order to "fix" such resource
> false‑positive, I do quick dis/enable the resource or as in
> this very instance rpm updates which restarted node.
> Again, how cluster might think resource is okey while
> debug‑monitor shows it's not.
> I only do not know how to reproduce this in a controlled,
> orderly manner.
Maybe try "xentop" or "watch virsh list" on all nodes and watch what is
> thanks, L
> Manage your subscription:
> ClusterLabs home: https://www.clusterlabs.org/
More information about the Users