[PATCH] Increase writetimeupdatedelay and sharedelay in Samba4.pm, remove winbind from Samba for 4.3

Michael Adam obnox at samba.org
Tue Jun 30 01:03:15 MDT 2015


failed again - resubmitted

I get the impression that we should disable the defer_open
test against the ad_dc_ntvfs env in order to be able
complete any autobuild ... :-/


On 2015-06-30 at 00:21 +0200, Michael Adam wrote:
> On 2015-06-29 at 10:53 +0200, Michael Adam wrote:
> > On 2015-06-29 at 09:27 +0200, Stefan (metze) Metzmacher wrote:
> > > Hi Andrew,
> > > 
> > > >> What this patch does is remove the *internal* winbind from
> > > >> the source4 directory.  This was replaced by default with
> > > >> winbindd from source3 in Samba 4.2, and this patch is to
> > > >> remove it.  Unlike other areas of duplication, this has no
> > > >> redeeming features in my view, so now we have fixed the
> > > >> classicupgrade issues (on which this patch builds), we can
> > > >> remove it. 
> > > >>
> > > >> I'm running private autobuilds to confirm I got the
> > > >> knownfail changes correct, but in the meantime I want to
> > > >> ensure there is the opportunity for reasonable discussion
> > > >> here.
> > > > 
> > > > The attached two patches pass autobuild on the Catalyst Cloud
> > > > other than the usual knownfail on the notify tests.  To do
> > > > that, I had to match the writetimeupdatedelay and sharedelay
> > > > parameters in Samba4.pm to match Samba3.pm.  I'm not sure
> > > > what about the different winbind implementations changes the
> > > > timing in the ntvfs file server, but this seems a reasonable
> > > > change, if only for consistency. 
> > > > 
> > > > Now I have that settled, could someone please review/push?
> > > 
> > > This causes a lot of flakey tests, mainly because the
> > > server_maxtime is reached.
> > > 
> > > We should try to remove some tests, source3/selftest/tests.py
> > > seems to call a lot of rpc. tests also against ad_dc, which
> > > should already be covered by source4/selftest/tests.py. Can you
> > > have a look at that?
> > > 
> > > So in the meantime can someone please push the attached patch
> > > that changes the maximum runtime to 3 hours.
> > 
> > rbm, pushing.
> 
> FYI: autobuild failed three or more times on this push
> for me already with the below failure in the defer_open test.
> Retrying ...
> 
> Michael
> 
> > > But I also got this at least two times:
> > > 
> > > [1373(7743)/1865 at 1h40m14s]
> > > samba4.ntvfs.cifs.krb5.base.defer_open(ad_dc_ntvfs)
> > > smbtorture 4.3.0pre1-DEVELOPERBUILD
> > > Using seed 1435535709
> > > Starting 4 clients
> > > 4 clients started
> > > pid 11998: Testing deferred open requests.
> > > pid 11998: create[0,1]...
> > > ...
> > > pid 11999: open 3
> > > pid 11999: open 3 closed, status NT_STATUS_OK
> > > pid 11999: deferred test finished
> > > 
> > > UNEXPECTED(failure):
> > > samba4.ntvfs.cifs.krb5.base.defer_open.defer_open(ad_dc_ntvfs)
> > > REASON: Exception: Exception: child 3 (pid 12000) failed: Timing
> > > incorrect 0.00 violation 1 sec == 0.10
> > 
> > This and some kcc tests seem to be the most
> > prominent flakey tests at the moment, which
> > started last to happen much more frequently
> > again since June 24/25.
> > 
> > Cheers - Michael
> > 
> 
> 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20150630/94512d6b/attachment.pgp>


More information about the samba-technical mailing list