[ClusterLabs] Antw: Re: How to correctly stop cluster with active stonith watchdog?

Ulrich Windl Ulrich.Windl at rz.uni-regensburg.de
Mon May 6 03:03:53 EDT 2019


>>> Andrei Borzenkov <arvidjaar at gmail.com> schrieb am 05.05.2019 um 07:43 in
Nachricht <033573b9-188f-baf6-e4b9-ba73150a3e60 at gmail.com>:
> 30.04.2019 19:47, Олег Самойлов пишет:
>> 
>> 
>>> 30 апр. 2019 г., в 19:38, Andrei Borzenkov <arvidjaar at gmail.com>
>>> написал(а):
>>> 
>>> 30.04.2019 19:34, Олег Самойлов пишет:
>>>> 
>>>>> No. I simply want reliable way to shutdown the whole cluster
>>>>> (for maintenance).
>>>> 
>>>> Official way is `pcs cluster stop --all`.
>>> 
>>> pcs is just one of multiple high level tools. I am interested in 
>>> plumbing, not porcelain.
>> 
>> Yep, in such case pcs do ssh connection to all nodes to down
>> pacemaker and corosync.
>> 
>> Well, for mantaince you can look for clusterlab plumbing analog of
>> `pcs node standby`, this command doesn’t stop corosync and pacemaker,
>> but shutdown all resources on that node, so quorum will not lost.
> I appreciate the efforts, but the question was not "how to do
> maintenance" but "how to stop cluster". There are reasons to do it.
> 
> Consider emergency shutdown due to power outage. I definitely would not
> like nodes started to panic in the middle of shutdown.

It seems they simply forgot to implement such a command. AFAIR in HP-UX 
ServiceGuard the command was "cmhaltcl -f -v" ;-)
And it's NOT the same as when _trying_ to stop both nodes concurrently; it's a
clean shutdown!

Regards,
Ulrich


> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users 
> 
> ClusterLabs home: https://www.clusterlabs.org/ 





More information about the Users mailing list