Wednesday, July 6, 2016

[389-devel] Re: Replication after full online init

On Wed, 2016-07-06 at 17:38 +0200, Ludwig Krispenz wrote:
> Hi Noriko,
>
> I have test scenario for not correctly handled modrdns during total init:
>
> - have a database with n entries, n large enough tthat total init takes
> long enough to be able to apply an update while it is running
> - add two entries
> -- n+1: cn=child,$SUFFIX
> -- n+2: cn=parent,$SUFFIX
> both have the same parentid and n+2 will be replayed after n+1
> - start total update
> - do a modrdn
> cn=child,$SUFFIX
> changetype: modrdn
> newrdn: cn=child
> newsuperior: cn=parent,$SUFFIX
>
> now cn=child,cn=paren,$SUFFIX will be sent before its parent
>
> I do not know how we can fix this, I think it is a corner case, but we
> should keep it in mind
>

Wow, excellent find!

What is the actual cause of this? It sounds like a race condition
between the sort and the transmission of the entries ....

--
Sincerely,

William Brown
Software Engineer
Red Hat, Brisbane

No comments:

Post a Comment