Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 2004-06-19T23:11:51,
Eugene Crosser <crosser at rol.ru> said:
> > The current CVS version is unstable and corrupts the data; this is
> > reproducible on our test setups too, and on Philipp's also. It's a work
> > in progress right now - hopefully fixed RSN ;-)
> With CVS version of Friday morning, I cannot reproduce corruption with
> significant activity, and several node switchovers. Good!
Very good to know!
> 1. After manual "drbdadm disconnect", *first* subsequent attempt to
> "connect" produces stack trace and device stays "StandAlone". Second
> attempt to "connect" succeeds and apparently no other harm is done.
> Maybe it's triggered by the fact that both notes where made primary (to
> check data integrity), and then the "wrong" one was manually made
> secondary before trying to connect them again.
"Both nodes primary" is an absolute error condition, basically. (To
check integrity, you should have went secondary/secondary and then
compared md5sums of the backing store.)
> Second thing: when I did "invalidate all" there was a lot of messages
> line these (some characters where lost on serial console):
New territory, probably not much used yet.
In any case, your bugs are valid, however I'm happy that all they
amounted to is a) in unlikely paths, and b) annoying log messages and no
data corruption, not even a crash. That's not quite perfect yet, but
already rather close and not a real show-stopper.
Mit freundlichen Grüßen,
Lars Marowsky-Brée <lmb at suse.de>
--
High Availability & Clustering \ ever tried. ever failed. no matter.
SUSE Labs | try again. fail again. fail better.
Research & Development, SUSE LINUX AG \ -- Samuel Beckett