[ClusterLabs] Antw: Re: DLM fencing

[Digimer]

On 10/02/16 02:40 AM, Ulrich Windl wrote:
>>>> Digimer <lists at alteeve.ca> schrieb am 08.02.2016 um 20:03 in Nachricht
> <56B8E68A.1060301 at alteeve.ca>:
>> On 08/02/16 01:56 PM, Ferenc Wágner wrote:
>>> Ken Gaillot <kgaillot at redhat.com> writes:
>>>
>>>> On 02/07/2016 12:21 AM, G Spot wrote:
>>>>
>>>>> Thanks for your response, am using ocf:pacemaker:controld resource
>>>>> agent and stonith-enabled=false do I need to configure stonith device
>>>>> to make this work?
>>>>
>>>> Correct. DLM requires access to fencing.
>>>
>>> I've ment to explore this connection for long, but never found much
>>> useful material on the subject.  How does DLM fencing fit into the
>>> modern Pacemaker architecture?  Fencing is a confusing topic in itself
>>> already (fence_legacy, fence_pcmk, stonith, stonithd, stonith_admin),
>>> then dlm_controld can use dlm_stonith to proxy fencing requests to
>>> Pacemaker, and it becomes hopeless... :)
>>>
>>> I'd be grateful for a pointer to a good overview document, or a quick
>>> sketch if you can spare the time.  To invoke some concrete questions:
>>> When does DLM fence a node?  Is it necessary only when there's no
>>> resource manager running on the cluster?  Does it matter whether
>>> dlm_controld is run as a standalone daemon or as a controld resource?
>>> Wouldn't Pacemaker fence a failing node itself all the same?  Or is
>>> dlm_stonith for the case when only the stonithd component of Pacemaker
>>> is active somehow?
>>
>> DLM is a thing onto itself, and some tools like gfs2 and clustered-lvm
>> use it to coordinate locking across the cluster. If a node drops out,
>> the cluster informs dlm and it blocks until the lost node is confirmed
>> fenced. Then it reaps the lost locks and recovery can begin.
>>
>> If fencing fails or is not configured, DLM never unblocks and anything
>> using it is left hung (by design, better to hang than risk corruption).
>>
>> One of many reasons why fencing is critical.
> 
> I'm not deeply in DLM, but it seems to me DLM can run standalone, or in the
> cluster infrastructure (we only use it inside the cluster). When running
> standalone, it makes sense that DLM has ist own fencing, but when running
> inside the cluster infrastructure, I'd expect tha tthe cluster's fencing
> mechanisms are used (maybe just because if the better logging of reasons).

To be clear; DLM does NOT have it's own fencing. It relies on the
cluster's fencing.

-- 
Digimer
Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without
access to education?