[PATCH] mark tombstone_reanimation patch flapping
kamenim at samba.org
Fri Mar 20 15:43:46 MDT 2015
On Fri, Mar 20, 2015 at 8:18 PM, Andrew Bartlett <abartlet at samba.org> wrote:
> On Fri, 2015-03-20 at 14:40 +0200, Kamen Mazdrashki wrote:
> > Hi Andrew,
> > Sorry for delay, I didn't do anything yesterday - just fall asleep all
> of a
> > sudden.
> > I think both delete_object and tombstone_reanimation tests interfere with
> > replication.
> > As far as I can recall, the idea for all those tests was to pause
> > replication during
> > the test. We haven't had implementation in place to pause replication
> > then
> > and imo we don't have it now too.
> Yes, replication can be stopped, but can you give a theory as to how
> replication causes an issue in tombstone_reanimation?
No clear view at the moment - only theory I have is this bug:
Pointed out by Mark few months ago (thanks Mark)
> The delete_object test does manipulate replication, indeed, that is the
> purpose of the test.
Well, that's embarrassing - I thought we don't have implementation in place
to pause replication during the test, but apparently we do... implemented
> The dbcheck failures I wasn't previously aware of - it is one thing to
> have the wrong state after a conflict in delete or not, it is entirely
> another to get that state wrong enough that dbcheck can notice. This
> may give some insight.
It seems I have missed that dbcheck fails
Do you have logs for that?
> Andrew Bartlett
> Andrew Bartlett http://samba.org/~abartlet/
> Authentication Developer, Samba Team http://samba.org
> Samba Developer, Catalyst IT
More information about the samba-technical