[DRBD-user] primary inconsistent

maurizio oggiano oggiano.maurizio at gmail.com
Wed Oct 14 15:11:53 CEST 2009

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

below you can find some piece of information about my problem:

> On Tue, Oct 13, 2009 at 11:54:54AM +0200, maurizio oggiano wrote:
> > Hi all,
> > I have a problem with a cluster degraded with a node inconsistent state.
> > Is there a way to recovery a consistent state even if the sync source is dead?
> > Maybe by means a snapshot, but I don't know how to do that.
> Can you elaborte on the situation?

Yes. I have two nodes A e B syncronized.
- A dies
- B becomes primary and  an application does  some change on data
managed by drbd
- A reboots, so a syncronization starts from B to A.
- the syncronization is interrupetd beacuse B dies. In this case I
have a degraded cluster with a node with inconsistent data so It can't
perform his functions

> Is this just a "what iff" question?
>  --> we provide the before-sync-target and after-sync-target hooks.
> if you use them as intended, you always have something consistent
> to fall back to.

In this case I want apply a strategy to recover a snapshot of a
consistent data in order to guarantee  ( even if a loss of data may
happen) that degraded cluster works with consistent data. My question
is the following : Is drbd is capable to take a snapshot of consistent
data and use it in case of a degraded primary with inconsistent data?
o maybe have I to write some script to use with  before-sync-target
and after-sync-target hooks?
In this case if i take a snapshot when "before-sync-target"  handler
is activated can data managed by drbd change during the time I'm
taking a snapshot?

> Are you already in a situation where your
> only remaining copy of data had been sync target before?
> Then you are in the unlucky situation, that your data is inconsistent.
> Depending on how far the resync progressed before, and how desperate you
> are in trying to reconstruct some of the content,
> and of course, how much file system meta data blocks would have been
> target of resync, you may be able to still get to most of the files,
> in some state or other. But I'd probably rather go for the latest backup
> than trying to do cumbersome forensics on a half-way bulk-updated file
> system image.
> Did I misunderstand the question?


More information about the drbd-user mailing list