[ClusterLabs] Antw: Re: Antw: Re: Antw: [EXT] Re: Q: constrain or delay "probes"?

Ulrich Windl Ulrich.Windl at rz.uni-regensburg.de
Mon Mar 8 08:35:36 EST 2021


>>> Andrei Borzenkov <arvidjaar at gmail.com> schrieb am 08.03.2021 um 11:46 in
Nachricht <366c7071-8b7e-9ea8-5ea1-cbb6de6d4e4f at gmail.com>:
...
> Probe needs to answer "is resource active *now*". If probe for resource
> is impossible until some other resources are active, something is really
> wrong with design. Either resource is active or not.

The resource state could actually be one of starting, started, stopping, and stopped.
Independent of the actual state is whether you can detect the state.
(Adding complex frameworks can make this task quite hard it seems)
"virsh list" which ist used to get the state relies on libvirtd (it seems to me). As libvirtd is a a super-daemon (can do "everything"), it also relies on the locking daemon, locking daemon needs a common filesystem and so on (as I indicated).
One could use an external NFS server to provide the locking space, but that's still another external dependency that could
halt your cluster...
...

Regards,
Ulrich





More information about the Users mailing list