gluster_community_weekly_meeting
LOGS
12:00:25 <atinm> #startmeeting Gluster community weekly meeting
12:00:25 <zodbot> Meeting started Wed Jan 13 12:00:25 2016 UTC.  The chair is atinm. Information about MeetBot at http://wiki.debian.org/MeetBot.
12:00:25 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
12:00:25 <zodbot> The meeting name has been set to 'gluster_community_weekly_meeting'
12:00:33 <atinm> Greetings everyone
12:00:44 <atinm> Its time for this week's community meeting
12:00:58 <atinm> #info agenda is right here at https://public.pad.fsfe.org/p/gluster-community-meetings
12:01:16 <atinm> #topic roll call
12:01:28 <atinm> Who all do we have here today?
12:02:00 * obnox waves
12:02:09 * obnox kinda here.. :-)
12:02:21 * jdarcy is
12:02:31 <atinm> Welcome obnox, jdarcy
12:03:00 <atinm> Lets wait for couple of minutes to get into the real action :)
12:03:24 <atinm> Appreciate if people can vote their presence
12:03:49 <atinm> There is no point of having a hour long community meeting with three people :/
12:04:08 * ndevos _o/
12:04:14 <obnox> 4
12:04:22 <obnox> ndevos: \o/
12:04:29 <atinm> Still a less number to me :(
12:04:48 <obnox> atinm: absolutely, especially if one (me) is not to be counted fully
12:04:56 * raghug here
12:04:56 <obnox> (when it comes to core gluster business)
12:05:12 <obnox> 5
12:05:51 <atinm> jdarcy, ndevos, obnox, raghug : Should we really continue with this number?
12:06:17 * sankarshan_ lurks
12:06:19 <jdarcy> I'd be OK with collecting AI status via email/etherpad instead.
12:06:45 <atinm> Allright
12:06:59 <atinm> lets make it a very short then
12:07:08 <atinm> #topic AIs from last meeting
12:07:29 <atinm> ndevos, do you have any updates on the reminder mail on maintainer list about active backports?
12:07:55 <atinm> ndevos, IIRC I've not seen it till now
12:08:29 <atinm> ndevos, I take it as No
12:08:33 <atinm> #action ndevos to send out a reminder to the maintainers about more actively enforcing backports of bugfixes (this year)
12:08:40 <atinm> rastar and msvbhat to consolidate and publish testing matrix on gluster.org. amye can help post Jan 1.
12:08:59 <atinm> rastar, msvbhat : any updates from amye ?
12:09:31 * ndevos just got a call, sorry
12:09:41 <atinm> I am skipping the AIs where the representatives are not in for today
12:09:59 <atinm> ndevos, no issues, I've carried it forward
12:10:08 <atinm> #action kshlm & csim to set up faux/pseudo user email for gerrit, bugzilla, github (after csim comes back on 4th Jan)
12:10:17 <atinm> #action rastar and msvbhat to consolidate and publish testing matrix on gluster.org. amye can help post Jan 1
12:10:34 <csim> well, i am back :)
12:10:57 <atinm> #action kkeithley to send a mail about using sanity checker tools in the codebase
12:11:15 <atinm> csim, ahh!! so you have any updates on that AI?
12:11:33 * anoopcs arrives late.
12:11:40 <atinm> #action kshlm to write up a README for glusterfs-specs
12:11:51 <jdarcy> Oh, oh, I get one!
12:12:08 <csim> atinm: I do not remember being contacted for it, but maybe I forgot
12:12:10 <jdarcy> I wrote up something, submitted it as a patch, I think there's general agreement.  Just needs to be merged.
12:12:54 <atinm> csim, no issues, I've carried it forward, we could check the status on next week then
12:13:01 <ndevos> yeah, I know Humble has merge powers, but he seems to be ignoring irc (and isnt in this channel0
12:13:57 <atinm> Humble is out for some visa work
12:13:59 <jdarcy> OK, I'll bug him as well.
12:14:09 <atinm> So moving to next one
12:14:15 <ndevos> #link http://review.gluster.org/13187
12:15:10 <atinm> #topic GlusterFS 3.7
12:15:33 <atinm> pranithk : are you around?
12:16:08 <atinm> Right time to join pranithk
12:16:17 <atinm> We were asking about 3.7 updates
12:16:40 <atinm> Any plans for the next release pranithk ?
12:16:41 <pranithk> atinm: 3.7.7. is held on NetBSD regressions for the patches that need to be merged.
12:17:11 <atinm> #info 3.7.7. is held on NetBSD regressions for the patches that need to be merged
12:17:22 <atinm> pranithk, so do you have any deadlines in your mind?
12:17:54 <pranithk> atinm: if netbsd regressions pass in the next hour I will release in 2 hours. So as soon as it completes I am ready to release it.
12:18:04 <atinm> pranithk, :)
12:18:32 <atinm> #info 3.7.7 will be release once all the blocker patches pass NetBSD regression runs and are merged
12:18:51 <atinm> pranithk, anything else you want to share?
12:18:57 <pranithk> atinm: nope
12:19:03 <atinm> pranithk, thanks
12:19:06 <atinm> Moving on
12:19:38 <atinm> #topic GlusterFS 3.6
12:20:03 <atinm> #info 3.6.8 is been released, however the official announcement hasn't happened yet I believe
12:20:13 <atinm> So I am marking an AI on kshlm
12:20:26 <atinm> #action kshlm to announce the availability of 3.6.8
12:20:39 <atinm> Any other points on 3.6 front?
12:21:06 <atinm> I take it as 'No' and moving on
12:21:08 <raghu> atinm: nope. I am planning to make 3.6.9 on 20th Jan
12:21:30 <atinm> #info 3.6.9 is planned on 20th Jan 2016
12:21:37 <atinm> #topic GlusterFS 3.5
12:21:52 <atinm> ndevos, time for you to speak :)
12:22:10 <ndevos> was planning to do a release this weekend...
12:22:25 <ndevos> but the not starting of regression tests kept me busy :-/
12:22:53 <ndevos> there are some new patches (all related to quota) sent, maybe I include those as well
12:23:10 <atinm> ndevos, so end of this week?
12:23:12 <ndevos> that would make 3 patches in total for the next release, maybe this week
12:23:40 <atinm> ndevos, is the next version 3.5.8?
12:23:43 <ndevos> schedule says "around the 10th of each month", this week counts as "around"
12:24:03 <ndevos> yes, I think 3.5.8, but would need to check to be sure
12:24:24 <atinm> #info 3.5.8 will be released by end of this week
12:24:28 <ndevos> 3.5.8 it is
12:24:34 <atinm> cool
12:24:46 <atinm> ndevos, anything else on 3.5?
12:24:50 <ndevos> nope
12:25:04 <atinm> ndevos, thanks for the updates
12:25:06 <atinm> Moving on
12:25:14 <atinm> #topic GlusterFS 4.0
12:25:23 <atinm> jdarcy, your turn?
12:25:35 <atinm> I'd add about GD2 later on
12:25:35 <jdarcy> OK...
12:25:58 <jdarcy> Still working on reconciliation.  Decided to scrap the old code, after spending days on it.  :(
12:26:14 <jdarcy> Haven't heard much from DHT-land lately.
12:26:58 <jdarcy> One idea has been discussed, of making NetBSD/FreeBSD support official 4.0 features.
12:27:03 <atinm> do we have Shyam or Venky around?
12:27:36 <jdarcy> Apparently we have some (funded!) FreeBSD folks who might be able to drive that.
12:27:57 <atinm> excellent news!
12:28:11 <ndevos> that sounds great, Manu could use some assistence :)
12:28:25 <csim> \o/
12:28:30 <csim> funded by whom ?
12:29:15 <jdarcy> Their company, iX Systems I think.
12:29:48 <ndevos> there was an email in december from them, they were trying to get in touch about 'offical' FreeBSD support
12:30:38 <atinm> I am extracting some data from the trello board (would be visible publicly shortly) for DHT2 ongoing work
12:31:12 <jdarcy> Next AI?
12:31:14 <atinm> DHT2 is currently working on the POC for different fops like readv, (f)sync, (f)truncate, statfs, xattrs etc
12:31:23 <jdarcy> Ah, OK.
12:31:30 <csim> ndevos: we have 1 single freebsd slaves for now :/
12:31:47 <atinm> From GlusterD 2.0 side we have started working on etcd bootstrapping
12:32:26 <atinm> kshlm is still focusing on stabilizing the txn framework
12:32:31 <ndevos> csim: it is only used for (short) smoke tests now, and the email (to the -devel list?) mentioned that they might be able to help out with testing and things - lets see where it goes
12:32:35 <jdarcy> Yay etcd bootstrapping.
12:33:00 <jdarcy> What should we do with the plugin proposal?  Fold it into the GlusterD2 doc (where?) or publish separately?
12:33:05 <atinm> jdarcy, how about the plugin proposal? Are you going to open it up to the  community?
12:33:15 <atinm> jdarcy, :)
12:33:29 <atinm> jdarcy, Do you want to put it in our design wiki?
12:33:54 <atinm> jdarcy, as thats the place holder for all our design docs as of now and later on we can move all of them to gluster-specs
12:34:01 <jdarcy> There's a design wiki?
12:34:08 <atinm> jdarcy, yes
12:34:24 <ndevos> oh, where is that?
12:34:26 <atinm> jdarcy, https://github.com/gluster/glusterd2/wiki/Design
12:35:26 <ndevos> but, but, but, gluster-specs has more eyes on it than a wiki nobody gets notifications about?
12:36:00 <atinm> ndevos, I agree
12:36:07 <ndevos> it would really be nice to all features to follow the same workflow...
12:36:20 <atinm> ndevos, as I said we would need to move it to spec
12:36:39 <atinm> ndevos, let me try this out, in another week's time let me move them to spec
12:37:01 <atinm> #action atinm to move all the design docs of GD2 to glusterfs-specs
12:37:07 <jdarcy> :)
12:37:09 <ndevos> atinm: what is blocking from getting it done before next meeting?
12:38:05 <atinm> ndevos, 1 more week could also be before next week's meeting :)
12:38:28 <atinm> We also do have a 4.0 roadmap page which is live now
12:38:34 <atinm> #info 4.0 roadmap page :https://www.gluster.org/community/roadmap/4.0/
12:39:00 <ndevos> atinm: "could", but maybe you can #action it for yourself or one of the other GD2 devs?
12:39:07 <atinm> aravindavk, Do you want to update anything on the eventing side?
12:39:40 <aravindavk> design discussion is in progress. Shyam and Luis provided inputs for design. Working on it
12:39:48 <atinm> aravindavk, cool
12:40:05 <atinm> ndevos, I will ensure that you wouldn't complaint in next week's meeting :D
12:40:22 <ndevos> atinm: ok, thank you :)
12:40:41 <atinm> I believe that ends all of our 4.0 discussions
12:40:46 <atinm> jdarcy, anything to add?
12:41:05 <jdarcy> Nope.
12:41:22 <atinm> #topic Open Floor
12:41:40 <atinm> Do we have anything to discuss here?
12:41:59 <ndevos> maybe about the current failing of regression tests?
12:42:13 <atinm> I'd like to remind all our developers about the change in regression trigger
12:42:46 <jdarcy> Good idea.
12:42:52 <atinm> Please refer to http://www.gluster.org/pipermail/gluster-devel/2016-January/047752.html
12:43:23 <rjoseph> Do we have a conclusion on the current NetBSD regressions?
12:43:51 <obnox> as I have written in my mail, I think it is really really sad, if automatic regressions are disabled
12:43:58 <ndevos> we also have failing tests due to an improvement in the regression job, if tests fail with a Java stack trace, check http://thread.gmane.org/gmane.comp.file-systems.gluster.infra/775
12:44:54 <atinm> ndevos, Also I have seeing jobs getting aborted very frequently, slaves are getting offline in between
12:45:07 <jdarcy> rjoseph: That's part of the rationale for making NetBSD a 4.0 feature.  Until then, regressions are run *post* merge, and issues fixed on a best-effort or fixed-resource basis.
12:45:08 <atinm> s/have/have been
12:45:22 <obnox> my comment was regarding the regression trigger
12:45:25 <rastar> obnox: by automatic do you mean trigger on patch update?
12:45:30 <jdarcy> I've aborted a lot of NetBSD runs that were hung for 12+ hours.
12:45:42 <obnox> rastar: yes, my ideal workflow would be this:
12:45:47 <jdarcy> Not today, but a bunch last week.
12:45:56 <atinm> jdarcy, rastar changed it in a way that no job should wait for more than 5 hours
12:46:13 <rjoseph> jdarcy: that's great
12:46:15 <atinm> I think we are discussing too many things at one go
12:46:19 <rastar> jdarcy: atinm there is timeout enabled for runs300 mins
12:46:19 <jdarcy> atinm: Do we have a way to track how often that "kill switch" triggers?
12:46:46 <ndevos> obnox: I'm a little sad about that too, but I hope at least a +1 code-review from anyone can get the tests started
12:47:01 <obnox> rastar: dev commits patch -- some or all tests are run automatically -- only if tests pass reviewers start reviewing -- if reviews are granted then it can be mergede, but this would force rebase (ffw) and re-run of a regression test if a rebase was necessary
12:47:16 <ndevos> the ugly solution to +1 verify what people are doing now, defeats the change
12:47:23 <atinm> jdarcy, not that I am aware of
12:48:39 <atinm> ndevos, but don't you think verified +1 means that patch has been verified and tested by the author and that should give some confidence to the reviewers?
12:49:11 <atinm> ndevos, as I mentioned in my reply that I was in favour of triggering regression on a +1, but I am just mapping it with a +1 verified, am I wrong?
12:49:11 <obnox> atinm: right, but 'some confidence' is poor compared to full passed regression runs or even build and smoke
12:49:22 <ndevos> atinm: not if developers know it will trigger a regression test
12:49:25 <rjoseph> if the regression is automatically triggered after +1 verify then the reviewer should start review after the regression result.
12:49:30 <atinm> obnox, on +1 verified the regression is run
12:49:45 <obnox> atinm: that does not seem to make a lot of sense to me
12:50:02 <jdarcy> Maybe we need +1/+2 Verify.
12:50:04 <obnox> verified sounds kind of synonymous to regressions have been run
12:50:05 <ndevos> atinm: it also makes it impossible to see in the listing-view if regressions did a +1, or if it was the patch author
12:50:51 <atinm> ndevos, no that's not right
12:51:01 <ndevos> I would like to have different labels for different tests, Smoke=+1, NetBSD=+1, Linux=+1 etc..
12:51:04 <rastar> obnox: my version of workflow
12:51:15 <atinm> ndevos, if its a regression vote it will be +1ed by either Gluster or NetBSD build system
12:51:31 <rastar> obnox: dev sends patch, immedietely sets +1 verified, regressions pass, reviewers review, patch is merged
12:51:42 <atinm> ndevos, if its by something else then that name would be there
12:51:56 <ndevos> atinm: yes, but how do you list patches that have been +1'd by regression tests? I need to open each change before I can see those details
12:52:10 <rastar> obnox: even in this workflow, reviewers review only after full confidence of regressions being complete
12:52:15 <obnox> rastar: fine. so with verify+1 you have the 'old' mode. that is good
12:52:22 <rastar> obnox: yes
12:52:25 <jdarcy> Definitely, something that's visible in a listing is preferable to something that's only visible in the per-patch view.
12:52:31 <obnox> and omitting +1 you can send WIP patches or so?
12:53:06 <obnox> rastar: but then, the final merge happens after possible rebase and regression-re-run ?
12:53:30 <atinm> ndevos, I never thought about this point
12:53:35 <ndevos> a Gerrit admin should be able to create additional labels, and users should only be able to set Verified, where as automated tests can then set their own label
12:53:38 <atinm> ndevos, that's valid
12:53:51 <rastar> obnox: yes for ommiting +1 for WIP
12:54:11 <obnox> rastar: and merge only with rebase and (if it WAS rebased) a new regression run?
12:54:11 <rastar> obnox: tests just before merge is not possible without zuul
12:54:23 <obnox> rastar: zuul ?
12:54:45 <rastar> obnox: yes that is gate keeping tool for merging patches
12:55:04 <obnox> hm
12:55:10 <rastar> obnox: http://docs.openstack.org/infra/zuul/
12:55:28 <rastar> obnox: specifically written by openstack guys when they had same problem as ours
12:55:31 <obnox> well, samba has the home-grown autobuild. that also does it
12:55:38 <rastar> and found gerrit and jenkins insufficient
12:55:38 <jdarcy> Since we're almost out of time, perhaps we need to move this back to email.
12:55:45 <atinm> rastar, but ndevos brings a valid point in terms of listing down the patches which has been +1'd by regressions
12:55:45 <rjoseph> rastar: After each patch update will the regression retrigger? Currently all acks go away after patch update
12:55:48 <obnox> we _only_ have autbuild and not the whole rest
12:56:17 <atinm> we are running out of time
12:56:32 <atinm> Can we continue this discussion on the mail thread?
12:56:35 <obnox> samba also found gerrit insufficient. for this reason and because it can't properly handle patch_sets_m but that is another story
12:56:40 <rastar> ndevos: atinm this works? status:open reviewer:"Gluster Build System <jenkins@build.gluster.com>"
12:56:41 <obnox> yeah mail thread
12:56:45 <obnox> atinm: ^^
12:56:55 <rastar> ok, mail thread
12:57:02 <obnox> atinm: but sometimes such discussion is faster
12:57:14 <obnox> (than reading whole novels in mail threads :-)
12:57:26 <ndevos> rastar: doenst work when verified=-1 is set, or someone else did a verified=+1
12:57:32 * rjoseph agree with obnox
12:57:52 * obnox running for next meeting
12:57:57 * obnox waves
12:58:01 <rastar> ndevos: ok will look into it
12:58:09 <rastar> lets discuss in mail
12:58:15 <atinm> rastar, thanks
12:58:17 * rastar should go for next meeting too
12:58:28 <ndevos> rastar: having different labels for different tests will also prevent any tests from overwriting results of other tests
12:58:46 <rastar> ndevos: I agree, kshlm wanted to try that out
12:58:46 <ndevos> rastar: will you start an email about the label things?
12:58:53 <rastar> ndevos: sure
12:58:56 <jdarcy> ndevos: I would like to pursue that idea.  Or, more precisely, I'd like someone to.  ;)
12:59:02 <ndevos> rastar: thanks!
12:59:18 <ndevos> jdarcy: we'll need to poke kshlm a little for that, I think :)
13:00:22 <atinm> Allright guys, that brings to the end of today's meeting and it was quite a productive one especially at the end, thanks all for joining, will see you next week at same time
13:00:44 <atinm> #endmeeting