12:09:37 <kshlm> #startmeeting Weekly Community Meeting 06/Jul/2016
12:09:37 <zodbot> Meeting started Wed Jul  6 12:09:37 2016 UTC.  The chair is kshlm. Information about MeetBot at http://wiki.debian.org/MeetBot.
12:09:37 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
12:09:37 <zodbot> The meeting name has been set to 'weekly_community_meeting_06/jul/2016'
12:10:31 <kshlm> #topic Roll Call
12:10:41 <post-factum> o/
12:10:41 * ndevos _o/
12:10:43 * jdarcy o/
12:11:27 <post-factum> annnd that's all?
12:11:29 <ndevos> hey, its at least 4 of us
12:11:56 <kshlm> post-factum, It's a holiday in India today
12:12:00 <ndevos> it is a holiday in Bangalore, many developers will be offline
12:12:06 <post-factum> oh, I see
12:12:37 <ndevos> only a select few is *really* dedicated
12:12:46 <ndevos> .. few = 1
12:13:18 <jdarcy> It's both Idul Fitr and Ratha Yatra.  Interesting.
12:13:22 <post-factum> what kind of holiday?
12:13:40 <kshlm> post-factum, What jdarcy said
12:14:12 <kshlm> With the 4 of us, shold we continue with the meeting?
12:14:41 <jdarcy> Ah, Idul Fitr == Eid al-Fitr
12:14:47 <ndevos> maybe just quick status updates for the meeting minutes?
12:14:50 <jdarcy> I'm willing to give it a pass.
12:15:11 <post-factum> i have several notes on 3.7, but one should be addressed to skoduri, and one to you, kshlm
12:15:20 <post-factum> and that's all
12:15:28 * msvbhat arrives bit late
12:15:33 <post-factum> 5!
12:15:42 <kshlm> Okay, let's have a quick status update and end this meeting quickly.
12:16:08 <kshlm> I'll start with 3.7 since theres more interest in it.
12:16:12 <kshlm> #topic GlusterFS-3.7
12:16:21 <post-factum> #link http://review.gluster.org/#/c/14254/
12:16:38 <kshlm> We had a gfapi breakage in 3.7.12 which we've identified, and have a fix.
12:16:48 <post-factum> I made dumb port of it to 3.7, but self-heal daemon fails now
12:16:49 <kshlm> But we need to first get the fix into master.
12:16:54 <jdarcy> See also: http://review.gluster.org/#/c/14866/
12:17:11 <ndevos> teh fix was waiting on a test-case, I think
12:17:16 <ndevos> is that still the case?
12:17:52 <kshlm> ndevos, Last I heard from rastar was that the exiting gfapi tests already had a test which shold have caught this.
12:17:57 <kshlm> But they don't run right now.
12:18:05 * kkeithley arrives late :-(
12:18:16 <kshlm> rastar said he'd comment on the review with this information.
12:18:24 <ndevos> kshlm: hmm, I did not see that note anywhere...
12:18:50 <kshlm> I asked him to do it probably late last evening.
12:19:23 <kshlm> We'll need to wait till tomorrow then.
12:19:31 <ndevos> yes, I guess so too :-/
12:19:41 * partner too..
12:19:57 <ndevos> it's easy enough to write a test case for it, but if they already have it, I'm not going to write a new one
12:20:13 <kshlm> post-factum, The review you linked, is that the reconnection thing that we chatted about some time back?
12:20:21 <post-factum> kshlm: correct
12:20:30 <post-factum> kshlm: it breaks shd for 3.7, and i dropped it
12:20:44 <post-factum> kshlm: with "transport.address-family not specified. Could not guess default value from (remote-host:(null) or transport.unix.connect-path:(null)) options" error
12:21:02 <kshlm> Good to know. I've not tested that change out completely yet.
12:21:10 <kshlm> I'll keep this in mind.
12:21:12 <partner> kshlm: so if using libgfapi, no updating into 3.7.12 as of now? (sorry, i'm slow reading backlog)
12:21:33 <kshlm> partner, Yep. Please don't.
12:21:45 <partner> thanks, i was just about to, already announced..
12:21:58 <post-factum> partner: or cherry-pick several commits
12:22:09 <ndevos> partner: qemu+gfapi is a no-go, with some corner cases, other libgfapi applications seem to work fine
12:22:18 <kshlm> Not yet announced, we've been busy figuring out what went wrong.
12:23:03 <post-factum> isn't http://review.gluster.org/#/c/14854/ the final fix?
12:23:08 <post-factum> for gfapi
12:23:14 <post-factum> +qemu
12:23:27 <kshlm> post-factum, It is.
12:23:31 <ndevos> post-factum: fix yes, patch no, it needs a test-case
12:23:45 <post-factum> also, http://review.gluster.org/#/c/14779/
12:23:55 <post-factum> ndevos: that is what i mean ;)
12:24:03 <kshlm> ndevos, I see a comment from rastar about the test case just before your -2.
12:25:41 <ndevos> post-factum: yeah :D but there is a test-case for that, but it should have been mentioned in the commit message
12:26:07 <ndevos> post-factum: the test-cases for gfapi can not be compiled/run in our regression-tests yet, so it has to be done manually
12:26:34 <ndevos> kshlm: yes, but his comment is not clear to me, *where* is he adding the test-case?
12:26:51 <kshlm> jdarcy, regarding 14866, is this something you're targetting for 3.7 as well?
12:27:16 <ndevos> kshlm: the "enable gfapi tests" patch needs much more work, I would not like to wait for that
12:27:36 <jdarcy> kshlm: Probably not, though there is an argument to be made that we shouldn't be running with completely unprotected GlusterD ports on any release.
12:27:58 <post-factum> kshlm: jdarcy: 14866 looks too big for fixes-only 3.7 branch
12:28:17 <kshlm> ndevos, so we just have a single gfapi test for this then?
12:28:39 <ndevos> kshlm: yes, or tell me what existing test-case covers it
12:28:49 <kshlm> ndevos, I have no idea.
12:29:06 <ndevos> kshlm: and so we wait until tomorrow :)
12:30:30 <kshlm> jdarcy, I agree with post-factum with this. Changes a lot how glusterd works.
12:30:43 <kshlm> And this will break rolling upgrades.
12:30:58 <jdarcy> True enough.
12:31:14 <post-factum> won't this break rolling upgrade from 3.7 to 3.8?
12:31:18 <ndevos> sounds like master/3.9 only then
12:31:30 <kshlm> post-factum, Any rolling-upgrade.
12:31:30 <post-factum> or to whatever branch this would be merged into
12:31:42 <post-factum> sounds very sad
12:32:17 <kshlm> post-factum, That is required to get secured glusterd by default though. Can't have it any other way.
12:32:45 <kshlm> jdarcy, So as this a 3.9 feature, we should be getting a feature-spec for it.
12:33:03 <jdarcy> Well, it might be possible for rolling upgrade to work if secure management is already enabled, but that would require extra work from the GlusterD maintainers.
12:33:53 <jdarcy> Also it would require fixing the "GFAPI never worked with secure management" bug, which also broke USS, etc.
12:34:24 <post-factum> so shouldn't extra efforts be commited into maintaining rolling upgrade possibility?
12:34:33 <kshlm> jdarcy, enabling secure management does require a downtime. Only people already running secure management could possibly do a rolling upgrade.
12:36:27 <kshlm> post-factum, enabling encrypted management/secure management is not possible without a downtime as of now.
12:36:30 <partner> could someone please throw me a link on what "secure management" exactly means? TLS enabled or something?
12:37:07 <kshlm> partner, That's it. TLS for glusterd, which we call management encryption right now.
12:37:25 <partner> rgr, thanks
12:37:44 <jdarcy> partner: That's pretty much it - TLS for GlusterD connections.  Really it has been unacceptable to do anything else for years now, but I didn't push on it too hard until recently.
12:38:19 <partner> jdarcy: this feature/better-ssl ?
12:39:00 <jdarcy> partner: Part of it, i.e. the first bullet.
12:39:49 <partner> nice, thanks, please proceed :)
12:39:55 <partner> with the meeting..
12:40:33 <kshlm> jdarcy, Now that the change out, I think we should at least announce it to get feedback.
12:40:51 <jdarcy> kshlm: Agreed.
12:41:10 <kshlm> From the comments here many people will be really interested in understanding how we will handle upgrades.
12:41:28 <kshlm> We can get this discussion happening now.
12:41:28 <post-factum> kshlm: precisely
12:42:20 <partner> indeed upgrades are always the interesting part.
12:43:03 <kshlm> jdarcy, Can you start this discussion on the mailing lists?
12:43:21 * ndevos is wondering how automated this all will be, or how it can be hooked in with FreeIPA, and how Kerberos mounts fit in there
12:43:28 <jdarcy> OK, but I'm really not sure if I'm the right person to be carrying this forward indefinitely.
12:44:37 <kshlm> I'll help as a GlusterD maintainer.
12:47:09 <kshlm> #action jdarcy will call for comments on the proposal to enable management encryption by default in 3.9. kshlm will help.
12:47:47 <kshlm> We should continue this discussion forward on the mailing lists.
12:48:42 <kshlm> We started with updates on 3.7 and did 3.9 as well.
12:49:07 <kshlm> Anyone else have updates or topics to discuss?
12:49:27 <post-factum> not now. need skoduri for that ;)
12:49:52 <kshlm> post-factum, :)
12:50:09 <kshlm> kkeithley, Since I took over todays meeting, will you be hosting the next meeting?
12:50:12 <kkeithley> sure
12:50:16 <kshlm> Thanks.
12:50:35 <kkeithley> provided my desktop doesn't lock up and take three reboots to get dropdown menus working
12:50:36 <kshlm> post-factum, Look who's just arrived.
12:50:39 <post-factum> :D
12:50:43 <post-factum> skoduri: welcome
12:51:14 <skoduri> post-factum, hi
12:51:41 <post-factum> skoduri: if you remember our last debugging session and have some time, i have an update for you, and we could proceed in #gluster-dev later
12:51:52 <skoduri> post-factum, oh sure
12:52:02 <post-factum> skoduri: ok will ping you after meeting
12:52:08 <skoduri> okay
12:52:29 <kkeithley> are we going to get 3.8.1 before ndevos gets pulled aside for a week ?
12:53:00 <ndevos> I plan so, yes
12:53:13 <ndevos> maybe tomorrow, or Friday
12:53:32 <partner> cool
12:53:39 <kshlm> I was planning to do 3.7.13 maybe tomorrow or Friday.
12:54:04 <ndevos> it'll be a joined release then :D
12:54:09 <kshlm> As both would be emergency releases, could we do it together?
12:54:41 <ndevos> yes, that should be fine, 3.8.1 should be done on Sunday, but I'll be travelling and no idea how busy I am next week
12:54:50 <kshlm> Joint release it is then.
12:55:14 <kkeithley> done on Sunday? Versus tomorrow or Friday?  Which is it?
12:55:26 <kshlm> I'll get rastar and poornimag to get on top the gfapi fixes first thing tomorrow.
12:56:03 <kkeithley> what about that other fix for 3.8?
12:56:12 <kkeithley> http://review.gluster.org/14859
12:56:29 <ndevos> Sunday is not possible, I'll be in a plane, so tomorrow or Friday for 3.8
12:56:48 <kshlm> kkeithley, That is the fix I was referring to.
12:56:55 <kkeithley> ah, "should be done on Sunday, i.e. 10th)
12:57:26 <kkeithley> IOW the normally scheduled date for a 3.8 release
12:58:18 <kshlm> We're 2 minutes from the end.
12:58:33 <kshlm> The very quick meeting turned out to last longer than expected.
12:58:48 <post-factum> hey, ppl joined, that is good :)
12:59:00 <kshlm> kkeithley, ndevos, Do you have any thing to share on Ganesha/Samba?
12:59:08 * ndevos not
12:59:19 <kkeithley> definitely nothing about Samba. ;-)
12:59:21 <kshlm> post-factum, 9 people here now I think.
12:59:24 <kkeithley> And nothing about Ganesha
12:59:28 <kshlm> Fine then.
12:59:39 <kshlm> This is the end of the meeting then.
12:59:54 <kkeithley> Samba is ira's and obnox's ball of wax
13:00:30 <kshlm> Thanks all for attending todays very irregular meeting.
13:00:35 <kshlm> #endmeeting