15:13:53 <kkeithley> #startmeeting Gluster Community Meeting
15:13:53 <zodbot> Meeting started Wed Jun 21 15:13:53 2017 UTC.  The chair is kkeithley. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:13:53 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:13:53 <zodbot> The meeting name has been set to 'gluster_community_meeting'
15:14:12 <kkeithley> #topic roll call
15:14:16 <kkeithley> who is here
15:14:47 <JoeJulian> here
15:15:20 * kkeithley wonders where ndevos went
15:15:49 <ndevos> oh, yes! I'm here
15:16:08 <jstrunk> here
15:16:31 * kkeithley wonders where amye is
15:17:12 <kkeithley> hmm, with this turnout I wonder if it's really worth continuing
15:17:41 <ndevos> amar has an item on the agenda, but he's missing too?
15:18:22 <kkeithley> indeed
15:18:54 <ndevos> maybe we can just go through the AI's that are listed?
15:19:21 <kkeithley> #topic AIs from last meeting
15:19:27 <kkeithley> ndevos will check with Eric Harney about the Openstack Gluster efforts
15:19:30 * shyam is here now!
15:20:07 <ndevos> oh, yes, that is still a todo - but more about finding a community member that would be willing to setup the test jobs
15:20:18 <kkeithley> and JoeJulian will share his conversations with Eric Harney
15:20:24 <ndevos> he did!
15:20:55 <kkeithley> oh, okay
15:21:23 <kkeithley> how about jdarcy will work with nigelb to make the infra for reverts easier
15:21:24 <ndevos> I cant find the email in the archives though :-(
15:21:50 <ndevos> no idea, is nigelb still around?
15:22:16 <kkeithley> rightl
15:22:44 <kkeithley> and certainly no joy for nigelb will document kkeithley’s build process for glusterfs packages
15:23:57 <ndevos> hmm :-/
15:24:26 <kkeithley> okay, well, we'll leave all those AIs open for next week then.
15:24:45 <ndevos> yeah, I guess so...
15:25:12 <kkeithley> Is anyone here to talk about Test cases contribution from community
15:25:15 <kkeithley> ?
15:25:29 <kkeithley> I guess that's more of the Eric Harney stuff?
15:25:53 <ndevos> not sure, there are little details about it in the agenda
15:26:36 <ndevos> the Eric Harney topic is for OpenStack Cinder test cases, I would expect this to be broader
15:26:53 <kkeithley> shyam: any status on 4.0 or 3.12?
15:27:08 <shyam> Nothing new other than what is in the page
15:27:19 <shyam> I will send out a nag for feature status as soon as 3.11.1 is tagged
15:27:55 <JoeJulian> ndevos: If I had even the slightest idea what was needed wrt the cinder topic I could decide whether or not I had the time to donate to that need.
15:28:06 <JoeJulian> EH won't answer.
15:29:10 <ndevos> JoeJulian: we'll need ot hook-up Jenkins with the openstack cinder Gerrit, so that each patch can get tested (with some Tempest configuration that uses a Gluster backend)
15:30:16 <JoeJulian> That doesn't sound too hard. Are there enough gluster owned resources for that?
15:31:10 <ndevos> JoeJulian: we can run it in the CentOS CI, that should have more than enough resources - its just people's time that is missing
15:32:57 <ndevos> I'll try to find some more links and details about it later this week or the next (just got back from 3 weeks travelling)
15:33:08 <kkeithley> anything else to say on this topic?
15:33:22 <ndevos> I dont think so
15:33:56 <kkeithley> #topic related projects
15:34:17 <kkeithley> nfs-ganesha 2.5.0 GA'd last week or so
15:34:23 <kkeithley> anything else of note?
15:34:58 <ndevos> gluster-swift is still a tough project to get in upstream distributions :-/
15:35:45 <kkeithley> because?
15:36:24 <ndevos> because it has a lot of OpenStack dependencies that are not part of Fedora or can easily be consumed by repositories in the CentOS Storage SIG
15:37:22 <ndevos> OpenStack is a very fast moving target, too fast for Fedora, it also makes maintaining gluster-swift a recurring excersise
15:38:16 <ndevos> not providing a good upstreamable project makes it more difficult to get adoption in the Gluster community (or elsewhere)
15:39:06 <ndevos> the recurring suggestions to look into other projects are Ceph RGW and Scalidity s3-server
15:39:14 <kkeithley> what ever happened to the idea of getting a good pluggable framework that would minimize what gluster-swift would need to add?
15:39:25 <ndevos> (and I'm not even alone anymore in suggesting the 1st :D)
15:40:08 <kkeithley> What about the Minio S3 server that FreeNAS uses?
15:40:09 <JoeJulian> There's also Harsha's project, minio.
15:40:46 <ndevos> it seems that OpenStack Swift does not like dropping all of its functionality to be replaced by gluster-swift - and the Swift3 layer on top of OpenStack Swift is not easily extendable either
15:41:05 <JoeJulian> Wow, you make it sound so foreign, kkeithley. :D
15:41:27 <kkeithley> really? o_O
15:41:31 <ndevos> yes, FreeNAS seems to use minio, but that does not have many features that we're looking for
15:41:47 <tdasilva> ndevos: i'm curious :D what problems are you running with swift?
15:42:05 <kkeithley> what features do we need that Minio doesn't support?
15:42:45 <ndevos> kkeithley: minio does not do multi-tenancy, it needs different servers for each tenant (from what people tell me)
15:42:59 <JoeJulian> Minio is AB and Harsha. I'm sure some Gluster love exists there that could be expanded upon.
15:43:49 <ndevos> tdasilva: I understood that user/group management is different in Swift vs S3, and that Swift3 can not handle it without issues
15:44:40 <ndevos> tdasilva: there are also other features that Swift does not provide but S3 users expect, something with bucket verioning and lifecycle stuff iirc
15:45:00 <ndevos> *bucket versioning
15:45:34 <tdasilva> ndevos: user/group mgmt is typically handled by auth mechanism, no? like keystone? I think bucket versioning is on the plans, lifecycle is a lot more difficult :/
15:46:09 <ndevos> minio was definitely one of the projects that some of us looked at, but it was deemed too far away from the features that we've got asked to provide
15:46:10 <tdasilva> ndevos: swift recently added a new bucket versioning mode to more closely match with S3s, so now the work needs to be done on the swift3 side
15:47:16 <kkeithley> what's the best guess for when that would get done?
15:47:44 <ndevos> tdasilva: Ram/Venkata has been trying to get in touch with the swift3 developers, but he was not very successfull... not sure how well swift3 is maintained
15:47:54 <tdasilva> kkeithley: honestly, I don't know...
15:48:11 <tdasilva> ndevos: #openstack-swift is a good place to find them
15:48:19 <tdasilva> look for timburke and kota
15:48:38 <ndevos> tdasilva: I expect he tried there, but I'll pass it along, thanks!
15:48:40 <tdasilva> and I'd be glad to help too :)
15:48:59 <ndevos> heh, of course, and ppai is assisting too
15:49:21 <JoeJulian> We could also invite harsha to the next community meeting to see if there's any thing he can add regarding minio.
15:49:43 <kkeithley> would you like an AI for that?
15:49:46 <JoeJulian> Sure
15:49:52 <tdasilva> kkeithley: https://review.openstack.org/#/q/status:open+project:openstack/swift3,n,z
15:50:11 <ndevos> I'll have to check who was looking at minio and see if they got in touch with the developers
15:50:18 <tdasilva> kkeithley: there's definetely versioning work going on,  bunch of patches that needs reviews...
15:50:21 <kkeithley> #action  JoeJulian to invite Harsha to next community meeting to discuss Minio
15:50:49 <kkeithley> #info https://review.openstack.org/#/q/status:open+project:openstack/swift3,n,z
15:50:57 <kkeithley> #info there's definetely versioning work going on,  bunch of patches that needs reviews...
15:51:00 <nigelb> kkeithley: The infra for reverts is done btw.
15:51:10 <nigelb> (sorry, I forgot we had a meeting today)
15:51:30 <kkeithley> #info The infra for simplified reverts is done btw.
15:52:06 <kkeithley> any other thoughts on S3/Swift?
15:52:36 <ndevos> not from me, not at the moment at least :)
15:52:38 <kkeithley> nigelb: any other infra updates?
15:53:06 <nigelb> we're planning for a gerrit upgrade on Monday.
15:53:10 <nigelb> Minor version upgrade
15:53:16 <nigelb> 2.12.2 -> 2.12.7
15:53:29 <nigelb> Not expected to cause major problems.
15:54:32 <kkeithley> #topic open floor
15:54:38 <kkeithley> anything else on anyone's mind?
15:54:49 <nigelb> This has been bouncing around in my head for a bit.
15:55:01 <ndevos> I wanted to get 3.8.13 released today, but I'll do it tomorrow
15:55:14 <nigelb> It's probably going to be controvertial, but I recommend we move all our docs to reference Centos as our preferred deployment platform.
15:55:28 <nigelb> Considering Fedora's release cycles, nobody is expected to run a production glusterfs setup on it.
15:55:39 <nigelb> Thoughts?
15:55:59 <kkeithley> I wasn't aware that we recommended anything at all about anything.
15:56:26 <kkeithley> I.e. it's free software, no warrantee, etc.
15:56:36 <ndevos> I was wondering about that too, I didnt think we recommended a certain distribution at all?
15:56:42 <nigelb> We sort of do.
15:56:44 <nigelb> And we sort of don't.
15:56:59 <nigelb> Our docs often refer to a mix of Fedora and CentOS.
15:57:18 <nigelb> Which looks a bit problematic because we refer to Fedora 18 in some places.
15:57:27 <kkeithley> lol
15:57:42 <nigelb> great way to make the documentation go out of date in 6 months :)
15:57:43 <kkeithley> maybe we have examples that say "this is how you do it on Fedora"
15:57:44 <ndevos> much of the docs is out-of-date...
15:57:55 <nigelb> Yes, there's an effort to fix that.
15:58:06 <kkeithley> I thought we had an effort underway to scrub the downstream docs and use them instead?
15:58:13 <nigelb> I sort of want to introduce this idea as a step in direction.
15:58:21 <nigelb> That's just for the admin guide.
15:58:27 <nigelb> (I believe)
15:58:49 <nigelb> But we'll still be ahead of the downstream docs in terms of features
15:59:13 <ndevos> I'm just a little biased, but I dont object to see pointers to the CentOS Storage SIG packages *everywhere* :D
15:59:21 <nigelb> Anyway, the point I'm trying to make is - By actually basing our docs on CentOS, we have a distinct advantage.
15:59:30 <nigelb> We run glusto tests agianst centos.
15:59:32 <nigelb> We know it works.
15:59:40 <shyam> +1 for CentOS as the recommendation or possibly the prime platform for testing etc. (which it already is)
15:59:51 <nigelb> So if the tests break, we know the docs need fixing.
16:00:07 <nigelb> at least the bits that are tested can be continually updated.
16:00:16 <shyam> CentOs Storage SIG for the packages not so much, considering the delay in packages appearing there... (sorry ndevos :) )
16:00:21 <kkeithley> IMO (IMHO) we need to be careful to avoid the appearance of recommending anything
16:00:38 <nigelb> I have a solution for the delay in packages, but I need some time to prep before I give you hope of that, ndevos
16:00:50 <nigelb> (I could just run glusto against the SIG pre-release packages)
16:01:15 <nigelb> kkeithley: The problem with that is we're in a state where our docs sort of don't work on anything.
16:01:35 <nigelb> If we say, we're going to make sure it works on one platform, at least we can consistently give a good UX.
16:01:35 <ndevos> yeah, if there are some automated tests against the packages, I'm happy to mark them for release quicker - now testing is only manual, and mainly by me whenever I have time
16:01:36 <shyam> kkeithley: Well yes, but used in all the examples in our documentation (this is how you would do X if it was CentOS and find a way in your favorite distro.)
16:02:23 <kkeithley> right, we can say things like "this is how to configure, e.g. for sytemd on CentOS.
16:02:30 <nigelb> Yes ^^^^
16:02:37 <nigelb> Exactly, what I'm trying to say.
16:02:42 <nigelb> No, "We prefer you use CentOS"
16:02:44 <kkeithley> But don't say "we recommend CentOS (because it's what we use here)"
16:02:53 <ndevos> "recommend" is probably too strong, just using CentOS in all examples would be good - and still allow contributors to extend the docs for other distributions
16:02:56 <nigelb> "We're using CentOS as our example"
16:03:02 <shyam> kkeithley: agreed
16:03:23 <nigelb> at the very least, everything we claim in our docs should work on CentOS.
16:03:34 <JoeJulian> Write instructions for what you know. If the community wants more, the community should provide.
16:03:36 <kkeithley> nigelb. Yes, that's fine
16:04:12 <nigelb> My primary motivation is so we can write tests that match our doc expectation.
16:04:34 <kkeithley> sounds good to me
16:06:07 <kkeithley> okay, we're over the hour. (even though we started 15 minutes late)
16:06:13 <kkeithley> anything else?
16:06:43 * ndevos does not have more
16:06:49 <kkeithley> motion to adjourn?
16:06:54 <kkeithley> going once?
16:07:02 <kkeithley> going twice?
16:07:14 <kkeithley> #endmeeting