Open Stack

----- Original Message -----
> On Wed, Mar 23, 2016 at 07:52:10PM +0000, Amrith Kumar wrote:
> > The meeting bot died during the meeting and therefore the logs on eavesdrop
> > are useless. So I've had to get "Old-Fashioned-Logs(tm)".
> > 
> > Action Items:
> > 
> > <amrith> #action [all] If you have a patch set that you intend to resume
> > work on, please put an update in it to that effect so we don't go abandon
> > it under you ...
> > <amrith> #action [all]  if any of the abandoned patches looks like
> > something you would like to pick up feel free
> > <cp16net> #action cp16net reply to trove-dashboard ML question for RC2
> > <amrith> #action [all] please review changes [3], [4], and link [5] in
> > agenda and update the reviews
> > 
> > Agreed:
> > 
> > <cp16net> #agreed flaper87 to WF+1 the patches in question [3] and [4]
> > 
> > Meeting agenda is at
> > https://wiki.openstack.org/wiki/Trove/MeetingAgendaHistory#Trove_Meeting.2C_March_23.2C_2016
> > 
> > Meeting minutes (complete transcript) is posted at
> > 
> > https://gist.github.com/amrith/5ce3e4a0311f2cc4044c
> 
> I'm still unsure of the value of adding these tests, and would love some
> pointers.
> 
> The current stable/liberty branch of trove fails
> gate-trove-scenario-functional-dsvm-mysql with a summary of "FAILED (SKIP=23,
> errors=13)"[1].
> 
> The 2 reviews in question fail with
>     "FAILED (SKIP=5, errors=2)"[2] and
>     "FAILED (SKIP=18, errors=1)"[3].
> 
> Granted these are heading in the right direction (in terms of fails).  There
> is
> no dependencies between the 2 reviews so I'll assume that if they're both
> merged that number wont regress.
> 
> If you look at the logs all 3 runs failed "instance_resize_flavor"
> [4][5][6]
> 
> So even with the changes merged you still don't end up with a gate job (even
> on
> the experimental queue that you can "just use".
> 
> Yes you have improved testing coverage, but is it meaningful?
> 
> I'm not blocking you from merging them I just don't understand the benefit.

The benefit is allow people using Liberty to have some meaningful results using them.

Without the fixes, you can't even run the basic instance creation tests, and you have broken code shipped with the release.
With them, you can run at least some basic tests, like instance creation, for datastores not supported by current set of integration tests.
Not all the tests run by the gate, but some, compared to the currently totally broken status.


-- 
Luigi