#opnfv-meeting: Weekly Technical Discussion
Meeting started by bh526r at 14:00:10 UTC
(full logs).
Meeting summary
- Roll Call (bh526r, 14:00:19)
- Bin Hu (bh526r,
14:00:34)
- Cristina Pauna (bh526r,
14:00:41)
- David McBride (bh526r,
14:03:31)
- Georg Kunz (bh526r,
14:05:16)
- Strategic Discussion - Code and Release Quality (bh526r, 14:06:55)
- Cristina shared the etherpad of draft proposal
of Code and Release Quality (bh526r,
14:07:53)
- Al Morton (bh526r,
14:08:08)
- The 1st is to improve project's info
accuracy (bh526r,
14:08:59)
- https://etherpad.opnfv.org/p/Code_and_Release_Quality
(bh526r,
14:09:51)
- Mark Beierl (mbeierl,
14:12:16)
- The identified problem is that inactive
committers had negative impact on progress of projects. An example
was Auto project. (bh526r,
14:15:07)
- Trevor Cooper (bh526r,
14:16:05)
- Proposal was made in etherpad (bh526r,
14:18:00)
- Bin mentioned that we can time this into
release cycle (bh526r,
14:18:54)
- David mentioned that we may automate Step B as
described in Etherpad (bh526r,
14:19:39)
- Al mentioned that we may set up a target date
for Step A, e.g. 6 weeks to act (bh526r,
14:20:23)
- David mentioned people may not be familiar with
how to add/remove committer. Cristina will provide link to wiki page
and example (bh526r,
14:21:41)
- Those suggestions were captured in
Etherpad (bh526r,
14:22:00)
- The 2nd is to improve Gerrit best
practices (bh526r,
14:22:43)
- i.e. committer not to merge own patch
(bh526r,
14:24:13)
- Suggestion about PTL exception and single
committer exception (bh526r,
14:28:44)
- Suggestion to start with A wiki page to
emphasize that self-merge is discouraged, and decide if we enforce
it later on. This will minimize the impact on projects that most
committers are inactive. (bh526r,
14:30:04)
- Suggestions are captured in etherpad
(bh526r,
14:30:44)
- The 3rd is to use FuncTest to verify installer
patches (bh526r,
14:31:15)
- This is already implemented. (bh526r,
14:32:00)
- The change is to document it and make it
mandatory in the future so that we won't drop this best
practice. (bh526r,
14:33:10)
- David asked if it is configurable. Cristina
suggested that if you want to disable it, it needs approval from
Release Manager or TSC (bh526r,
14:35:07)
- Suggestion is captured in etherpad (bh526r,
14:35:50)
- The 4th is to reduce the time it takes for
running the tests (bh526r,
14:36:22)
- The issue is that it takes long time to run all
tests in FuncTest test suites. It takes more than 4 hours in almost
all cases. On a regular basis, 10 hours even not enough (then timed
out after 10 hours) (bh526r,
14:38:33)
- Possible solution includes (A) parallel testing
(B) improve functest failure detection (C) daily jobs and weekly
jobs (bh526r,
14:40:23)
- Expect Cedric to drive the best solution
(bh526r,
14:40:44)
- David mentioned that we may run basic tests
initially, and advanced tests only after passing basic tests
(bh526r,
14:41:24)
- Cristina asked what is the reasonable time for
running FuncTest (bh526r,
14:43:34)
- Mark asked if the cycle time is for full test
or for a patch (bh526r,
14:44:04)
- Cristina clarified that it is not for a
patch (bh526r,
14:44:18)
- 2.5 years ago, it took 2.5 hours to run a full
cycle of functest (bh526r,
14:44:38)
- Mark mentioned that if one project is in the
formal process, 2-3 hours of running functest is reasonable on daily
basis. (bh526r,
14:47:46)
- David mentioned that it also intersects with
Release Process discussion (bh526r,
14:51:59)
- Test WG can continue working on it, while
Release WG keeps on working on process. (bh526r,
14:54:57)
- It is challenging to just say 20 minutes or 2
hours is enough (bh526r,
14:55:18)
- Cristina will send an email to mailing list,
with all comments we received today, and look for more input /
comments (bh526r,
14:56:24)
- The target is for TSC approval on Feb 5 (of at
least some of those ideas), if discussion is going well. Understand
some other ideas may need more time (bh526r,
14:58:12)
- AOB (bh526r, 15:02:29)
- Trevor C will give update if Infrastructure
comparison hosting v.s. service, and perhaps update of compliance
program (bh526r,
15:03:17)
- Meeting adjourned (bh526r,
15:03:23)
Meeting ended at 15:03:27 UTC
(full logs).
Action items
- (none)
People present (lines said)
- bh526r (48)
- collabot` (3)
- mbeierl (1)
Generated by MeetBot 0.1.4.