Re: ITS#7052, syncrepl, deletes, and MMR

[Howard Chu <hyc@symas.com>]

Rein Tollevik wrote:
> On 23.02.12 22:33, Quanah Gibson-Mount wrote:
> > --On Thursday, February 23, 2012 9:13 PM +0100 Rein Tollevik
> > <rein@OpenLDAP.org>  wrote:
> >
> > > The queuing of an old CSN done as a fix to ITS#7052 may have introduced a
> > > new race condition, an ITS and fix is coming.
> > >
> > > A new test script which exercise these race conditions is coming.
>
> > I was about to issue a testing call for RE24 for 2.4.30. If this ITS&
> > fix are going to be coming soon, I will hold off. Do you have an ETA on
> > them?
>
> After having analyzed my logs a bit more I'm fairly sure the 7052 fix
> didn't introduce any new races afterall :-)

No new races, but some old ones still remain. So far the only thing I've found 
to reliably prevent them is to serialize all writes in syncprov. Have not 
committed this yet; wondering if there's some other conditions we should place 
on this. E.g., only if mirrormode is configured?

--
  -- Howard Chu
  CTO, Symas Corp.           http://www.symas.com
  Director, Highland Sun     http://highlandsun.com/hyc/
  Chief Architect, OpenLDAP  http://www.openldap.org/project/