[DRBD-user] Choreography for db storage consistency check use case

Felix Frank ff at mpexnet.de
Thu Feb 2 13:46:47 CET 2012

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hi,

(going back on-list)

On 02/02/2012 12:31 PM, John Anthony wrote:
> The application is just a consistency check - so do not mind discarding
> the changes. 

you can do this then.

> Can I just disconnect the secondary and mount the drbd volumes even if
> that is a split brain, or do I have to go all the way and detach ? The

That doesn't compute. If you detach the local disk from a standalone
DRBD, you have no access to any data.

You have to disconnect and then cause split brain by promoting your new
StandAlone Secondary.

If you have a "snapshottable" backing device (e.g. LVM), you can use a
snapshot to even avoid split brain, but there are performance implications.

> thought is to try reduce the re-establish time to the extent possible!!

If you get the split-brain resolution right, re-establishing is about
instantaneous. Of course, the changes have to be undone by syncing back
from the Primary, but that happens transparently.

> Will give it a try today - meanwhile any ideas will be highly appreciated.

Needless to say: From the time of disconnection until everything is
finished and your nodes are back in sync, you loose all redundancy. If
the Primary dies now, you're screwed pretty thoroughly.

Cheers,
Felix



More information about the drbd-user mailing list