samba4.raw.notify.mask(dc) takes 97.668 seconds

[Volker Lendecke]

On Thu, Jan 03, 2013 at 09:38:40AM -0800, Matthieu Patou wrote:
> On 01/03/2013 12:45 AM, Volker Lendecke wrote:
> >On Thu, Jan 03, 2013 at 12:37:12AM -0800, Matthieu Patou wrote:
> >>Hi Volker,
> >>
> >>On 01/03/2013 12:33 AM, Volker Lendecke wrote:
> >>>On Wed, Jan 02, 2013 at 09:39:16PM -0800, Matthieu Patou wrote:
> >>>>Hi tridge,
> >>>>
> >>>>I was looking a bit more closely at the length of each testsuites (using
> >>>>the script from this change:
> >>>>http://git.samba.org/mat/?p=mat/samba.git;a=commit;h=5fe11f654ffbda913a03dec7045d850ca4866c2c)
> >>>>when I realize that the samba4.raw.notify took 104 seconds on my test 
> >>>>box
> >>>>which is quite ok (8 cores, lots of ram, ramdisk for the whole samba
> >>>>source
> >>>>tree, ...) and most of this is the mask test.
> >>>>
> >>>>Do you have any idea why is it so ?
> >>>Sure, this is because notifies are asynchronous. To verify
> >>>that something has not triggered you have to wait. The way
> >>>to improve time of this is to make the test parallel for the
> >>>different masks, probably via tevent_req jobs.
> >>Could it be possible also to reduce the wait delay ? as a plan B for
> >>reducing the length of this test ?
> >Possible, but this will make the test less reliable. If we
> >ever trigger notifies on something we should not, we will
> >not detect it under heavy load.
> Ok so appart from having tests from other environments running in 
> parallel the only other solution is to make this test more parallel.
> 
> Do you think that it's the same problem for the oplock tests ?
> 
> samba4.raw.lock(dc) -> 16sec
> samba4.raw.oplock(dc) -> 44sec

Yes, very similar.

Volker

-- 
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de