[DRBD-user] Automatic split brain recovery policies - after-sb-0pri - discard-older-primary
GAUTIER Hervé
herve.gautier at thalesgroup.com
Wed Jan 7 11:09:22 CET 2009
Florian Haas a écrit :
> On 01/06/2009 02:38 PM, GAUTIER Hervé wrote:
>
>> Hi there !
>>
>> After reading (one more time) the pretty DRBD user's guide
>> http://www.drbd.org/users-guide/s-configure-split-brain-behavior.html
>> I had noticed that the discard-older-primary policy for after-sb-0pri
>> was not listed.
>> I am not sure to have understood what does this policy, and even if it
>> still exists. Someone may explain me ?
>>
>
> Scroll down.
>
> "Note
>
> DRBD understands additional keywords for these three options, which have
> been omitted here because they are very rarely used. Refer to
> drbd.conf(5) for details on split brain recovery keywords not discussed
> here."
>
> I believe the drbd.conf man page has the information you're asking for.
> Does it not?
>
> Cheers,
> Florian
>
>
Shame on me ! RTFM ! Thank Florian !
Ok, but, excuse me, as my english is far from perfect, I don't
understand very well the discard-older-primary policy for after-sb-0pri:
"Auto sync from the node that became primary as second during the
split-brain situation."
Is that mean:
1 - the source of the synchronisation will not be the first node to be
primary after the split-brain detection, but the second ?
In case of split brain with this policy and so, when split brain has
just been detected and the resource is not in primary role on any host,
if a node becomes primary, any modification done will be lost,
and if this same node then becomes secondary, and another node becomes
primary, any modification will be kept for the sync.
Am I right ?
2 - the source of the synchronisation will be the first node to be
primary while the split-brain situation ?
That's the way I would have, as the discard-younger-primary policy does
not cover this case because it will select the node which assumed the
primary role last (before the split brain detection) as the source of
the synchronisation.
Thank in advance !
--
Hervé GAUTIER
More information about the drbd-user
mailing list