[DRBD-user] Fwd: confused with DRBD 9.0 and dual-primary, multi-primary, multi-secondary ...

[Lentes, Bernd]

Bernd Lentes

Anfang der weitergeleiteten Nachricht:

> Von: "Lentes, Bernd" <bernd.lentes at helmholtz-muenchen.de>
> Datum: 29. August 2018 um 23:32:49 MESZ
> An: Lars Ellenberg <lars.ellenberg at linbit.com>
> Betreff: Aw:⁨ [DRBD-user] confused with DRBD 9.0 and dual-primary, multi-primary, multi-secondary ...⁩
> 
> 
> 
> 
>>> Am 29.08.2018 um 16:47 schrieb Lars Ellenberg <lars.ellenberg at linbit.com>:
>>> 
>>> On Wed, Aug 29, 2018 at 03:38:34PM +0200, Lentes, Bernd wrote:
>>> Hi,
>>> 
>>> i found: 
>>> "With current DRBD-9.0 version running in Dual-Primary mode is not recommended (because of lack of testing).
>>> In DRBD-9.1 it will be possible to have more than two primaries at the same time."
>> 
>> The answer to your actual question is "Yes", below.
>> 
>> But read on, I try to give a more general answer for
>> "Can DRBD 9 do multi-primary?" first.
>> 
>> Even with DRBD 9, we can NOT do "multi-primary", but only "two-primary".
>> 
>> If you start playing around, it may at first *seem* like it would work.
>> It does not.
>> 
>> And *if* you do go "two-primary", that also means two nodes only.
>> 
>> NOT two-primary plus multiple secondaries.
>> Does not work either, unfortunately.
>> 
>> With the notable exception of live-migration in a multi-peer setup,
>> where we have "two" primaries with potentially "many" secondary peers.
>> But only for a short time, and these are not actually concurrently
>> written to, so that's a special case.
>> 
>> What I think we want to support at some point even in the "9.0" series
>> is two-primaries plus one or more secondaries. Maybe.
>> 
>> "N >= 3" Primaries is something that will come only
>> in some later "9.1" (or 9.2 or 10.x or 23.y) version.
>> If at all.  We don't see a "real" use case for this.
>> 
>>> Clear statement. I'm using SLES 12 SP3 with DRBD 9.0.8.
>> 
>> If you want DRBD 9, you should be prepared to keep track of,
>> and regularly upgrade to, upcoming DRBD 9 releases.
>> Currently we are at 9.0.15.
>> You should certainly not stay on 9.0.8.
>> 
>> If you really want/need "real, concurrent" dual-primary DRBD,
>> (which is most likely a bad idea anyways;
>> there are very limited use cases where dual-primary is actually useful
>> beyond being "cool" or apparently, but not actually, "convenient")
>> For now I'd suggest to stay with 8.4,
>> because that emphasizes the "two node only" aspect.
>> 
>> If you only need the "short time, not really concurrent" dual-primary
>> for "live migration" in a virtualization environment,
>> both DRBD 8.4 and DRBD 9 can do that,
>> and this is one of the main use cases for LINSTOR with DRBD 9.
>> 
>>> The kernel modules are 9.0.x.
>>> The drbd-utils i found are 9.x.
>>> 
>>> Are these already dual-primary capable ?
>> 
>> Yes.
>> 
>> But *really* at least upgrade to 9.0.15,
>> and then .16 and .17 and ... when they come out.
>> 
>> If that's not something you can do in your environment,
>> stay with the stable-and-frozen DRBD 8.4 series.
>> 
>> Or look into our LINSTOR project, which requires DRBD 9, and is supposed
>> to make integration with various virtualization solutions much easier.
>> Still, also in that case,
>> prepare to regularly upgrade both DRBD 9 and LINSTOR components.
>> There will be bugs, and bug fixes, and they will be relevant for your
>> environment.
>> 
>> -- 
>> : Lars Ellenberg
>> 
> 
> Hi Lars,
> 
> thanks for this very profound answer.
> 
> Bernd


Helmholtz Zentrum Muenchen
Deutsches Forschungszentrum fuer Gesundheit und Umwelt (GmbH)
Ingolstaedter Landstr. 1
85764 Neuherberg
www.helmholtz-muenchen.de
Aufsichtsratsvorsitzende: MinDir'in Baerbel Brumme-Bothe
Geschaeftsfuehrer: Prof. Dr. med. Dr. h. c. Matthias H. Tschoep, Heinrich Bassler, Dr. rer. nat. Alfons Enhsen
Registergericht: Amtsgericht Muenchen HRB 6466
USt-IdNr: DE 129521671
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20180830/76786095/attachment-0001.htm>