[ClusterLabs] Antw: Re: Antw: [EXT] Coming in Pacemaker 2.0.4: shutdown locks

[Andrei Borzenkov]

28.02.2020 01:55, Ken Gaillot пишет:
> On Thu, 2020-02-27 at 22:39 +0300, Andrei Borzenkov wrote:
>> 27.02.2020 20:54, Ken Gaillot пишет:
>>> On Thu, 2020-02-27 at 18:43 +0100, Jehan-Guillaume de Rorthais
>>> wrote:
>>>>>> Speaking about shutdown, what is the status of clean shutdown
>>>>>> of
>>>>>> the
>>>>>> cluster handled by Pacemaker? Currently, I advice to stop
>>>>>> resources
>>>>>> gracefully (eg. using pcs resource disable [...]) before
>>>>>> shutting
>>>>>> down each
>>>>>> nodes either by hand or using some higher level tool (eg. pcs
>>>>>> cluster stop
>>>>>> --all).  
>>>>>
>>>>> I'm not sure why that would be necessary. It should be
>>>>> perfectly
>>>>> fine
>>>>> to stop pacemaker in any order without disabling resources.
>>>>
>>>> Because resources might move around during the shutdown sequence.
>>>> It
>>>> might
>>>> not be desirable as some resource migration can be heavy, long,
>>>> interfere
>>>> with shutdown, etc. I'm pretty sure this has been discussed in
>>>> the
>>>> past.
>>>
>>> Ah, that makes sense, I hadn't thought about that.
>>
>> Is not it exactly what shutdown-lock does? It prevents resource
>> migration when stopping pacemaker so my expectation is that if we
>> stop
>> pacemaker on all nodes no resource is moved. Or what am I missing?
> 
> shutdown-lock would indeed handle this, if you want the behavior
> whenever any node is shut down. However for this purpose, I could see
> some users wanting the behavior when shutting down all nodes, but not
> when shutting down just one node.
> 

Well, this requires pacemaker supporting notion of "cluster wide
shutdown" in the first place.

> BTW if all nodes shut down, any shutdown locks are cleared.
> Practically, this is because they are stored in the CIB status section,
> which goes away with the cluster. Logically, I could see arguments for
> and against, but this makes sense.
> 

This actually allows poor man implementation of cluster wide shutdown by
setting lock immediately before stopping nodes; it could probably even
be integrated directly into "pcs cluster stop --all". I wish crmsh
offered something similar too.