[DRBD-user] Invalidate

Stephan Rattai srattai at zmnh.uni-hamburg.de
Thu Aug 4 14:47:44 CEST 2005

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


Am Donnerstag, 4. August 2005 14:38 schrieb Lars Marowsky-Bree:
> On 2005-08-04T14:35:42, Stephan Rattai <srattai at zmnh.uni-hamburg.de> wrote:
> > > it does not make sense to invalidate both.
> >
> > Even if I make a fool out of myself, I would really like to understand
> > why it doesn't make any sense, doesn't invalidate mean 'Hey, forget about
> > your metadata, you are inconsinstent, go get some good data' ?
>
> No, it means "Hey, your data is bad. get some new data from the peer." -
> if you tell both sides that they don't have good data, and that all
> their blocks are outdated, which one should drbd use? ;-) They are both
> bad!

Yeah, ok, I see that, I just didn't expect such an impact. Of course from 
drbd's point of view it's a mess. Next time I'll just zero the 
metadata-device if I ever want to start from scratch. 

> It's possible that this case hasn't been tested well, simply because no
> developer thought of trying it. (Which is usually a good indicator for a
> real bug.)
>
I guess it's impossible to get all the stupid ideas users would make :-)
>
> Sincerely,
>     Lars Marowsky-Brée <lmb at suse.de>



More information about the drbd-user mailing list