15:00:25 <hagarth> #startmeeting
15:00:25 <zodbot> Meeting started Wed Feb 12 15:00:25 2014 UTC.  The chair is hagarth. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:25 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:00:42 <hagarth> who do we have here today?
15:01:02 * jarrpa is here
15:01:08 * xavih is here
15:01:43 <hagarth> cool, let us get started. I suspect a few others are here but passive atm.
15:01:56 <hagarth> #topic AI follow up
15:02:24 <hagarth> gerrit for 3.3 and 3.4 has been sorted out - kkeithley is able to merge patches to those branches
15:02:43 * kkeithley is here
15:02:53 * ira is here
15:03:09 <hagarth> kkeithley has got samba packaging for CentOS - so we seem to be good there
15:03:27 <hagarth> jclift_: any progress on glupy renaming issue?
15:04:04 <hagarth> purpleidea did want some feedback on #link http://www.gluster.org/pipermail/gluster-users/2013-November/038050.html  from gluster developers - we seem to have missed that.
15:04:20 <hagarth> so let us carry forward the AI.
15:04:21 * jdarcy will do that today.
15:04:25 <hagarth> jdarcy: thanks!
15:04:53 <hagarth> #topic 3.5.0
15:05:16 <hagarth> we got beta3 out - geo-replication and quota seem to be mostly covered with this release
15:05:48 <hagarth> that has been a relief - we can get quite aggressive from here in getting upto 3.5.0 GA
15:06:20 <hagarth> johnmark has planned a test week for 3.5 starting today and we have a weekend test event in the Red Hat Bangalore office.
15:06:56 <hagarth> based on the outcome from these test fests, we should be able to release 3.5.0 soon
15:07:06 <jdarcy> W00t.
15:07:42 <hagarth> I will carve out more releases as and when patches get merged into release-3.5, so there might be a spike in the number of release candidates from here.
15:08:38 <hagarth> if you notice any blockers to 3.5.0 GA, please add it to the release tracker - #link https://bugzilla.redhat.com/show_bug.cgi?id=1049981
15:08:41 <glusterbot> Bug 1049981: unspecified, unspecified, ---, vbellur, NEW , 3.5.0 Tracker
15:09:00 <hagarth> any questions on 3.5.0?
15:09:25 <hagarth> guess not, let us move on
15:09:33 <hagarth> #topic 3.4.0
15:09:36 <kkeithley> http://review.gluster.org/6854
15:09:46 <kkeithley> we want this in 3.4 right?
15:10:08 <kkeithley> would someone review it so I can merge it and make a 3.4.0alpha1 release?
15:10:18 <kkeithley> 3.4.3alpha1 release
15:10:19 <hagarth> kkeithley: yes, that is right. It is a backport, I will do that after this meeting.
15:10:50 <hagarth> kkeithley: that would be great, let us get it going.
15:10:50 <kkeithley> I looked for it on master, but didn't see it. Not with that BZ
15:11:37 <kkeithley> commit b880b6b2908ad4e4afc8e26613bd0db8f0b28750,     BUG: 969461
15:11:43 <hagarth> kkeithley: this has possibly not made it to master in anticipation of the refactored AFR coming in
15:11:52 <hagarth> but this patch is present in release-3.5
15:12:05 <hagarth> and is essential for all releases that we support.
15:12:06 <johnmark> here
15:12:43 <hagarth> johnmark: cool, slight OT - would you be sending out details of test week for 3.5 today?
15:12:44 <kkeithley> okay, at this point I'll wait for the +2 review
15:12:51 <johnmark> hagarth: yup yup
15:12:52 <hagarth> kkeithley: ok
15:12:58 <johnmark> any particular pieces we shoudl focus on?
15:13:01 <kkeithley> I presme I have privs to set a tag in the release-3.4 branch
15:13:02 <johnmark> geo-rep is one
15:13:07 <kkeithley> presume
15:13:11 <hagarth> johnmark: quota would be another.
15:13:16 <johnmark> hagarth: ok
15:13:23 * johnmark goes to feature pages
15:13:24 <hagarth> kkeithley: if not, let us sort it over.
15:14:11 <hagarth> there is one more candidate for inclusion in 3.4.3
15:14:20 <hagarth> https://bugzilla.redhat.com/show_bug.cgi?id=1041109
15:14:22 <glusterbot> Bug 1041109: urgent, unspecified, ---, csaba, NEW , structure needs cleaning
15:14:41 <hagarth> we need to pay some attention to that.
15:14:52 <kkeithley> If I get 3.4.3alpha1 out — which I should — then we can test both
15:15:03 <johnmark> sweet
15:15:07 <hagarth> kkeithley: right
15:15:30 <hagarth> cool, anything else on 3.4?
15:15:46 <xavih> will this be included ? https://bugzilla.redhat.com/show_bug.cgi?id=859581
15:15:47 <glusterbot> Bug 859581: high, unspecified, ---, vsomyaju, ASSIGNED , self-heal process can sometimes create directories instead of symlinks for the root gfid file in .glusterfs
15:16:15 <johnmark> xavih: reproducible?
15:16:30 <xavih> johnmark: there is a test script to reproduce it
15:16:31 <kkeithley> xavih: it's in the 3.4.3 tracker
15:16:42 <xavih> kkeithley: ok
15:16:59 <hagarth> xavih: I think it would be good to review and get it into all branches - I see that you have submitted backports to all branches.
15:17:00 <kkeithley> may not get to it for 3.4.3alpha1 though
15:17:07 <xavih> there is a patch for review, but no reviews yet
15:17:31 <xavih> hagarth: yes
15:17:51 <hagarth> xavih: I will pick that up and nudge others  who are relevant to it.
15:18:00 <xavih> hagarth: ok, thanks
15:18:18 <hagarth> #action gluster devs to review patches for #link https://bugzilla.redhat.com/show_bug.cgi?id=859581
15:18:21 <glusterbot> Bug 859581: high, unspecified, ---, vsomyaju, ASSIGNED , self-heal process can sometimes create directories instead of symlinks for the root gfid file in .glusterfs
15:18:30 <hagarth> any other questions on 3.4?
15:18:46 <hagarth> guess not, let us move on.
15:18:55 <johnmark> :)
15:18:59 <hagarth> #topic 3.6
15:19:30 <hagarth> There have been a few features proposed over the last week (mostly carried over from 3.5 planning)
15:19:41 <hagarth> I also have got a few requests to move the date by a week
15:19:51 <hagarth> the original proposal freeze is tomorrow
15:20:01 <johnmark> hrm
15:20:07 <hagarth> if everybody feels it is ok, we can move the entire schedule by a week
15:20:12 <hagarth> thoughts?
15:20:12 <johnmark> hagarth: right
15:20:15 <johnmark> hagarth: +1
15:20:22 <johnmark> impossible to meet proposed deadline
15:20:27 <jdarcy> If we're going to move, I suggest two weeks.
15:20:40 <jdarcy> On the theory that you guys might come up with something in BLR.
15:20:45 <johnmark> jdarcy: good point
15:20:53 <johnmark> plus many US-based people are on break
15:21:02 <hagarth> jdarcy: ok, how about this - we move the schedule by 2 weeks
15:21:22 <hagarth> and have the proposal review meeting as part of this meeting 3 weeks from now?
15:21:44 <jdarcy> SGTM
15:22:01 <kkeithley> we get a break?
15:22:29 <johnmark> kkeithley: I mean, kids in school are out next week
15:22:33 <johnmark> doesn't pertain to you :)
15:22:44 <hagarth> #action hagarth to move schedule by 2 weeks and schedule a proposal review meeting for 3.6 on 5th march.
15:22:50 <jdarcy> I'm not exactly taking a break, but unavailable anyway due to FAST.
15:23:17 <hagarth> this will also give us enough buffer to push 3.5.0 out
15:23:48 <johnmark> cool
15:24:09 <hagarth> ok, any more questions on 3.6?
15:24:31 <hagarth> guess not, let us move on.
15:24:34 <hagarth> #topic 4.0
15:25:05 <hagarth> there have been some comments in the google doc page that jdarcy set up for 4.0
15:25:17 <hagarth> some of them call for more discussion
15:25:39 <hagarth> I wonder if we should pick up those from the doc and convert them to mailing list discussions
15:26:17 <hagarth> or if we should continue discussions in the google doc itself?
15:26:31 <jdarcy> ISTRC that Avati was amenable to creating a second draft as a way to kick off that discussion.
15:27:14 <hagarth> jdarcy: yeah, let me ping him on that one.
15:28:14 <hagarth> we have a lot of ground to cover there and need to evolve a good plan to develop 4.0 as a globally distributed community.
15:28:41 <jdarcy> Also from last week's mtg, there are some new pieces (esp. around scalability) that need to be incorporated.
15:29:20 <hagarth> jdarcy: right, let us try to get a v2 of the document out that incorporates all these comments.
15:29:35 <hagarth> and that can be the basis for further discussions on 4.0.
15:30:06 <hagarth> any topics related to 4.0?
15:30:32 <hagarth> guess not, let us move on.
15:30:48 <hagarth> #topic discussion on xavih's proposal
15:30:56 <hagarth> #link http://lists.gnu.org/archive/html/gluster-devel/2014-02/msg00004.html
15:31:14 * purpleidea is joining late, sorry
15:31:20 <hagarth> xavih: would you mind summarizing the proposal briefly here for the benefit of folks who might have not read it?
15:31:28 <jdarcy> I have a patch that might relate to this.  http://review.gluster.org/#/c/6978/
15:32:11 <xavih> basically it's a way to give clients temporary full control over inodes, so that they do not need to constantly send requests to the bricks
15:32:17 <jdarcy> It's only a tiny bit of infrastructure, but might be useful for prototyping the consistency IPC.
15:32:35 <hagarth> jdarcy: looks like an interesting idea to pursue
15:32:39 <johnmark> interesting...
15:32:58 <xavih> this also substitute the inodelk and entrylk fops
15:33:10 <jdarcy> One possibility is to use it for "long polling" to get server-to-client notifications.
15:33:18 <xavih> and allows a more aggressive cache implementation on client side that will significantly improve performance
15:33:50 <hagarth> jdarcy: right
15:34:38 <hagarth> the server-to-client notifications could be used for invalidation of anything that the client exclusively thinks it has control over
15:34:44 <xavih> ndevos made me aware that there is something very similar in NFSv4 called delegations and callbacks
15:35:00 <jdarcy> Maybe we should schedule a separate IRC meeting to discuss ideas esp. around how to handle failures in such a system.
15:35:12 <ira> xavih: SMB has similar concepts...
15:36:13 <hagarth> jdarcy: sounds like a good idea .. any preferred day and time for that meeting?
15:36:32 <jdarcy> This time tomorrow?
15:36:43 <ira> Should work.
15:36:45 <johnmark> ira: so something that all distributed FS try ot implement?
15:36:55 <xavih> ok
15:37:00 <hagarth> jdarcy: I have a conflict, but I think I can be on here too.
15:37:08 <ira> johnmark: Everyone has a way to do client side caching of various forms.
15:37:12 <jdarcy> johnmark: Yeah, it was common even when we did it in MPFS (1999-2000 or so).
15:37:13 <johnmark> jdarcy: is there a way for us to do this such that we are enabling or setting a standard for how this should be done in all DFS?
15:37:36 <johnmark> maybe that's a question for later, after we actually have an implementation :)
15:37:48 <ira> johnmark: Actually, that's a question, I'm looking at some...
15:37:59 <johnmark> ira: ah, ok
15:38:03 <hagarth> johnmark: there already are implementations - we might borrow a lot of ideas from all places :)
15:38:11 <ira> johnmark: Though really in the NFS/SMB space, clearly, it goes to all spaces.
15:38:11 <jdarcy> I think each DFS tends to have enough different data objects on which this must operate, and different semantics across those objects, that it would be hard to standardize.
15:38:12 <johnmark> hagarth: of course :)
15:38:26 <johnmark> jdarcy: that's the answer I was looking for
15:38:46 <ira> jdarcy: You'll never standardize it across Gluster and Ceph.  But within Gluster...
15:38:51 <jdarcy> But absolutely, we can and should borrow ideas from those (sometimes ourselves) who have gone before.
15:39:13 <ira> Even that standardization is a large effort.
15:39:36 <ira> To make sure, you can support all the strange semantics, people will want :/
15:40:04 <jdarcy> So, here 10am EST tomorrow?  At least Xavi, Ira, Vijay seem interested.
15:40:23 <kkeithley> zimbra invite?
15:40:24 <hagarth> jdarcy: sounds good to me
15:40:29 <johnmark> I would love to be here to breathe in your air, but I will unfortunately be heading to the airport for vacation :(
15:40:35 <jdarcy> kkeithley: I can do that.  You want in?
15:40:42 <kkeithley> sure
15:40:43 <purpleidea> johnmark: vacation!
15:40:48 <johnmark> purpleidea: I know, right?
15:41:07 <hagarth> jdarcy: we could also CC gluster-devel as an optional invitee
15:41:08 <purpleidea> johnmark: you need to send a fax first (remember?)
15:41:16 * jdarcy actually volunteered for two AIs.  Mark the date.
15:41:17 <johnmark> purpleidea: yes. I remember :)
15:41:20 <johnmark> lol
15:41:30 <hagarth> jdarcy: lol
15:41:48 <hagarth> ok, let us move on
15:41:55 <hagarth> #topic open discussion
15:41:59 <hagarth> 1. GSOC
15:42:11 <hagarth> vipulnayyar: are you around?
15:42:19 <purpleidea> 2. One question about if there isn't a better way to handle: https://gluster.org/pipermail/gluster-users/2014-February/039039.html
15:42:30 <purpleidea> (maybe in new style replication)
15:42:41 <hagarth> we have 2 days, 3 hours remaining for the organization deadline
15:42:41 <vipulnayyar> yes I'm here.
15:42:59 <johnmark> vipulnayyar: welcome
15:43:02 <hagarth> johnmark: can you help us with meeting the organization deadline?
15:43:24 <johnmark> hagarth: yes. pretty sure we can come up with two or three solid ideas and submit gluster.org as an organization
15:43:37 <johnmark> hagarth: I will take care of that before I leave
15:43:45 <johnmark> ...along with purpleidea's hotel fax :)
15:43:54 <hagarth> johnmark: great
15:44:14 <hagarth> vipulnayyar: do feel free to propose glusteriostat - we will provide any necessary assistance.
15:44:21 <purpleidea> hagarth: we should submit puppet-gluster of a GSOC and i'll apply :P
15:44:43 <jdarcy> purpleidea: Not sure if NSR would be any different.  It's a transport phenomenon, supposed to balance between fast failure detection and false alarms due to mere slowness.
15:44:47 <hagarth> johnmark: you can pick up ideas from our backlog
15:45:14 <vipulnayyar> Thanks hagarth .
15:45:15 <purpleidea> jdarcy: okay, well for whichever replication, a solution for that problem would be appreciated for sysadmins, i think.
15:45:35 * jdarcy will reply further on the ML.
15:45:36 <hagarth> one question on the trello backlog - I am planning to make it visible to everybody with edit rights to org 'GlusterFS'. is everybody fine with that?
15:45:45 <jdarcy> +1
15:46:27 <hagarth> done - #link https://trello.com/b/mczS4DKw/glusterfs-backlog is now world readable.
15:46:39 <hagarth> one less AI to carry further ;)
15:46:53 <johnmark> hagarth: ok
15:47:12 <johnmark> hagarth: +1
15:47:24 <hagarth> I will also send a note to mailing lists about our backlog.
15:47:31 <johnmark> vipulnayyar: I'll create a page for us ot submit ideas
15:47:36 <johnmark> hagarth: thanks
15:47:50 <vipulnayyar> sure, thanks johnmark
15:48:03 <hagarth> any more questions on GSoC?
15:48:24 <johnmark> nope. pretty straightforward
15:48:30 <hagarth> ok, moving on to Storage SIG in CentOS
15:48:46 <hagarth> johnmark: would you want to provide some context on this one?
15:49:20 <johnmark> hagarth: sure
15:49:22 <johnmark> one sec
15:49:45 <johnmark> CentOS has the concept of "SIG"s == special interest groups
15:49:58 <johnmark> basically, it's a way of providing a shared pool of packages for other projects to make use of
15:50:21 <kkeithley> Like EPEL, but more focused?
15:50:44 <johnmark> so, for example, there's a cloud SIG, and there are some base packages not in regular CentOS that some projects need, such as openstack, euca, et al
15:50:52 <johnmark> kkeithley: right
15:51:07 <johnmark> so the ceph project started the storage SIG to do the same thing
15:51:29 <johnmark> and we should probably participate to make sure that packages on CentOS can be used for full GlusterFS features
15:51:43 <johnmark> for example, we need an updated libvirt and qemu
15:51:55 <johnmark> and a samba compiled with our vfs plugin
15:52:01 <johnmark> so those sorts of things
15:52:09 <hagarth> johnmark: maybe G4S, pmux et al too
15:52:25 <purpleidea> hagarth: johnmark: *cough* and a puppet-gluster package...
15:52:43 <johnmark> you can see the storage SIG proposal here: http://grokbase.com/t/centos/centos-devel/141nvt6rde/storage-sig-proposal
15:52:44 <hagarth> purpleidea: I left out puppet-gluster intentionally ;)
15:52:52 <johnmark> purpleidea: lol... yes! of course
15:52:57 <purpleidea> hehe
15:53:01 <hagarth> johnmark: should we have a ML discussion to arrive at the list of packages?
15:53:11 <johnmark> hagarth: right - except those would not be essential for all storage build
15:53:18 <hagarth> johnmark: ok
15:53:22 <johnmark> those would be specific ot a gluster distribution, which we can submit at any time
15:53:31 <kkeithley> I'm happy, e.g. to do things like build gluster, samba4+gfapi, ganesha+gfapi, for the SIG, but what's the first step?
15:53:39 <hagarth> johnmark:  ok, got it.
15:53:47 <johnmark> but read through the storage SIG proposal, and it would be great if one of our devs could respond to that thread
15:53:58 <johnmark> or QE folks...
15:54:06 <johnmark> doh... where's Lala?
15:54:13 <hagarth> johnmark: lalatenduM is interested - he could not join this meeting today - I will sound him tomorrow.
15:54:19 <johnmark> hagarth: thank you
15:54:41 <hagarth> #action hagarth to talk to lalatenduM about storage SIG
15:55:07 <hagarth> kkeithley: I will ask lalatenduM to mark you when he responds.
15:55:25 <hagarth> anything more on this topic?
15:55:28 <johnmark> kkeithley: that does bring up the question of what a gluster distro looks like, however
15:55:38 <johnmark> for 3.5, what all do we want to include in the builds?
15:55:46 <hagarth> johnmark: we probably should propose a gluster SIG ;)
15:55:51 <johnmark> can we include g4s, pmux, ghadoop, puppet-gluster, etc?
15:55:59 <johnmark> hagarth: haha... perhaps :)
15:56:07 <purpleidea> johnmark: ^glupy ?
15:56:12 <kkeithley> I'd like to get g4s into Fedora first
15:56:13 <johnmark> but I think they'll just ask us to submit those project
15:56:18 <johnmark> purpleidea: yup, another one
15:56:28 <johnmark> er... submit those packages individually, which is fine
15:56:48 <hagarth> johnmark: it would be great to have a single bundle of all reasonably mature projects from forge
15:56:59 <johnmark> hagarth: we can aat least start the conversation onthe mailing list like this "we have a bunch of packages we want to put into centos - how can we do that?"
15:57:07 <johnmark> hagarth: exactly
15:57:21 <purpleidea> (there is also copr, which is a good place for new packages)
15:57:23 <johnmark> and it would be good to start getting them into other distros, as well
15:57:23 <hagarth> johnmark: yeah, let us start that - would you take the lead on that?
15:57:28 <hagarth> johnmark: right
15:57:42 <johnmark> hagarth: indeed. I've been talkign to the centos guys about it anyway - just need to push something on the list
15:57:46 <kkeithley> I guess I should whine on devel@fedoraproject.org about the unresponsive reviewer
15:57:51 <hagarth> johnmark: cool!
15:57:52 <johnmark> purpleidea: copr?
15:57:56 <johnmark> lol
15:58:13 <purpleidea> johnmark: https://fedorahosted.org/copr/
15:58:18 <hagarth> ok - one more topic for today.
15:58:22 <purpleidea> like ppa's for fedora
15:58:33 <hagarth> 3. 3.3 EOL ?
15:58:42 <purpleidea> +1
15:58:57 <jdarcy> +1 for -1
15:59:07 <hagarth> should we EOL support for 3.3 once 3.5.0 is out?
15:59:41 <kkeithley> if it means I can duck out of 3.3.3 then +1
15:59:49 <jdarcy> Absolutely.  Supporting old/current/new is quite enough IMO.
16:00:03 <hagarth> kkeithley: :)
16:00:28 <hagarth> ok, I think we have our answer. 3.3 to be EOL once 3.5.0 is out.
16:00:37 <hagarth> anything else for today?
16:01:04 <hagarth> figure not, thanks everybody and talk to you all next week.
16:01:07 <hagarth> #endmeeting