samba-tool drs replicate improvements

[Andrew Bartlett]

On Mon, 2016-08-01 at 21:17 +1200, Andrew Bartlett wrote:
> On Sat, 2016-07-30 at 07:33 +1200, Andrew Bartlett wrote:
> > 
> > On Thu, 2016-07-28 at 08:47 +0200, Stefan Metzmacher wrote:
> > > 
> > > 
> > > Hi Andrew and Garming,
> > > 
> > > can I get a review on the following patches.
> > > 
> > > The main change is a timeout of 5 minutes for DsReplicaSync()
> > > instead of 1 minute.
> > > 
> > > This may help with the flakey tests.
> > > 
> > > As a bonus I added a --async-op option so that
> > > we can also trigger a replication without waiting for the result.
> > > 
> > > Thanks!
> > > metze
> > 
> > I really do appreciate you looking into this.
> > 
> > Reviewed-by: Andrew Bartlett <abartlet at samba.org>
> > 
> > I'm sorry I've not started my further investigation, I got
> > sidetracked
> > closing bugs :-)
> 
> I've started my investigation, and so far I am still setting the
> timeouts.  I've got a pretty reliable way to replicate the issues
> (and
> it turns out ctdb issues!), so that is the first big step. 
> 
> I'm trying builds with some tracing logic specified, and this may
> help
> pin down the issue.

I'm pretty convinced this a not a real timeout, but a case of the error
case not triggering the IRPC (and then DCE/RPC) callbacks. 

In particular, when we see dsdb_replicated_objects_commit() fail in
dreplsrv_op_pull_source_apply_changes_trigger() because of an unknown
attid for a linked attribute), the failure does not propagate
correctly.

I'm also looking at better ways to handle
dsdb_replicated_objects_commit() failing. 

Andrew Bartlett

-- 
Andrew Bartlett                       http://samba.org/~abartlet/
Authentication Developer, Samba Team  http://samba.org
Samba Developer, Catalyst IT          http://catalyst.net.nz/services/samba