12:02:35 <hagarth> #startmeeting 3.8 planning
12:02:35 <zodbot> Meeting started Thu Dec  3 12:02:35 2015 UTC.  The chair is hagarth. Information about MeetBot at http://wiki.debian.org/MeetBot.
12:02:35 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
12:02:35 <zodbot> The meeting name has been set to '3.8_planning'
12:02:51 <hagarth> agenda - https://public.pad.fsfe.org/p/gluster-3.8-features
12:02:59 * jiffin is here
12:03:03 * itisravi is here
12:03:27 * ndevos _o/
12:03:31 * Manikandan is here
12:03:56 <hagarth> #topic Status Update on individual features
12:04:31 <hagarth> Leases Support for Gluster
12:04:46 * hgowtham is here
12:05:16 <hagarth> do we have any of the feature owners for this here? Soumya sent me an email update on this one
12:05:19 * sac is here
12:05:47 <ndevos> poornimag just joined after you mentioned the topic :)
12:05:48 * poornimag is late
12:05:49 <hagarth> poornimag: would you be able to provide an update on Leases?
12:05:56 <poornimag> sure,
12:06:20 * msvbhat is here
12:06:49 <poornimag> So the open item is that we need some effort from AFR team
12:06:59 <poornimag> apart from that everything is on track,
12:07:12 <hagarth> poornimag: what would the afr folks need to do?
12:07:21 * rjoseph is late
12:07:24 <poornimag> afr healing of leases
12:07:36 <poornimag> the design is in place,
12:07:54 <poornimag> if we cannot get this in for 3.8, it will be a tech preview release
12:07:56 * atinm is partially available
12:08:03 <hagarth> poornimag: ok
12:08:23 <hagarth> itisravi: any plans to add support for healing of leases in 3.8?
12:08:32 <poornimag> Sorry i missed, is there a dev freeze date set for 3.8?
12:08:41 <rjoseph> Right now we have dependency on AFR lock healing, Rebalance/DHT lock migration, Sharding fix for manatory locks
12:08:59 <itisravi> hagarth: we have listed it as an action item but there are no devs looking into it right now.
12:09:00 <rjoseph> lock migration and sharding fix is good to have
12:09:06 <ndevos> poornimag: if you mention "the design is in place", does that mean it is included in the glusterfs-specs repository?
12:09:08 <itisravi> its a stretch goal.
12:09:22 <hagarth> poornimag: not yet, we will be arriving at a schedule by the end of this meeting.
12:09:38 <poornimag> hagarth, oh ok
12:09:50 <hagarth> itisravi: ok..
12:10:05 * kshlm is late
12:10:45 <poornimag> ndevos, there were mails on devel regarding the same,
12:11:01 <rjoseph> ndevos: we have the initial design at glusterfs-specs
12:11:47 <hagarth> rjoseph: that sounds good
12:12:28 <spalai> sorry to interrupt. When is the dev-freeze for 3.8
12:12:55 <hagarth> spalai: we will be arriving at a schedule later in the meeting
12:12:59 <spalai> ok
12:13:08 <hagarth> anything else on leases?
12:13:12 <ndevos> poornimag: yeah, mails are good for the initial discussions, the result should be captured in a design doc before implementing things ):
12:13:18 <ndevos> uh, not ): but :)
12:13:29 <ndevos> thanks rjoseph
12:13:32 <rjoseph> ndevos: check the http://review.gluster.org/#/c/11980
12:13:36 * anoopcs arrives late
12:13:40 <ndevos> ok!
12:13:55 <hagarth> ok, let us move on to the next feature.
12:14:02 <hagarth> 3. Converged HA for NFS-Ganesha and Samba
12:14:19 <hagarth> kkeithley_: are you around for an update on this one?
12:14:42 <kkeithley_> no status, Jose Rivera is actually doing the work
12:14:58 * skoduri arrives late
12:14:58 <kkeithley_> Michael Adam or Ira maybe know something?
12:15:06 <ndevos> obnox: ?
12:16:05 <obnox> ndevos: pong
12:16:23 <ndevos> obnox: do you have a status update on "Converged HA for NFS-Ganesha and Samba" from Jose?
12:16:43 <ndevos> (this is the gluster 3.8 feature status meeting)
12:16:56 <obnox> ndevos: it is progressing.
12:17:15 <obnox> can get latest status from jose in a bit (1-2 hrs)
12:17:41 <hagarth> obnox: any guesstimates on when it might be available?
12:18:14 <obnox> hagarth: it depends on what available really means.
12:18:44 <obnox> i thing the upstream facing project should be consumable very soon. (few days..) but then workflows may need to be changed?
12:18:52 <obnox> consumption in downstream etc
12:19:17 <hagarth> obnox: ah ok, just trying to understand how every feature lines up with respect to a dev freeze upstream
12:19:41 <obnox> sorry for not being up to date - is this the gluster community meeting?
12:19:50 <obnox> (I thought it was wednesdays)
12:20:07 <hagarth> if dev completion is not possible within the next month or two, I would not aim to have that in 3.8.
12:20:24 <obnox> hagarth: oh, that should be very possible
12:20:28 <hagarth> obnox: this is a meeting on 3.8 feature status
12:20:35 <hagarth> obnox: sounds good to me!
12:21:07 <obnox> hagarth: Jose wanted to get it done during this month I think. ideally way before christmas break
12:21:18 <obnox> he also needs to concentrate on code smb stuff again :-)
12:22:01 <hagarth> obnox: that would be cool, we would need a feature page in our specs repository for this but so do many other features planned for 3.8. I will be sending reminders on that after this meeting.
12:22:19 <obnox> we will provide you with a more correct guesstimate later today
12:22:25 <hagarth> obnox: thank you!
12:22:44 <ndevos> obnox: where "you" is the upstream gluster-devel list? ;-)
12:22:52 <hagarth> moving on to features in quota
12:23:09 <vmallika> we are planning user and group quotas
12:23:10 <hagarth> vmallika, Manikandan: are you around?
12:23:44 <vmallika> we are preparing the design doc
12:23:47 <Manikandan> hagarth, yeah, we will update the same on devel mailing list and then proceed
12:23:53 <obnox> ndevos: dunno. if that's the place - sure I will ask jose to post there :-)
12:24:18 <vmallika> mid of next week we will share the design doc
12:24:21 <ndevos> obnox: thanks!
12:24:43 <hagarth> vmallika, Manikandan: thanks! I expect that we will have a better picture on implementation duration after the design is out?
12:25:05 <vmallika> yes
12:25:13 <Manikandan> hagarth, yes
12:25:35 <vmallika> we are concerned of performance impact when all 3 features are enabled dir, user and group quota
12:26:21 <hagarth> vmallika: right, nothing comes free of course. however we would need to measure the impact and keep it in acceptable limits.
12:26:37 <hagarth> vmallika, Manikandan: what about Enhance quota enable/disable in glusterd ?
12:27:33 <vmallika> we will update on this feature once the design for user quotas is done
12:28:02 <hagarth> vmallika: ok
12:28:10 <vmallika> we have planned how this to be enhanced, we will design and update on this soon
12:28:20 <Manikandan> We assume that should be comparatively easier
12:29:04 <hagarth> Manikandan: agree, let us discuss more once we have the design ready.
12:29:19 <hagarth> moving on from quotas
12:29:21 <Manikandan> hagarth, sure
12:29:28 <hagarth> next feature -  Provide mandatory lock support for Multiprotocol environment
12:29:56 <anoopcs> hagarth, Sent a mail to devel
12:30:00 <anoopcs> http://www.gluster.org/pipermail/gluster-devel/2015-December/047252.html
12:30:23 <anoopcs> hagarth, ^^
12:31:07 <hagarth> anoopcs: ok!
12:31:41 <hagarth> anoopcs: are there any open items in your design?
12:32:24 <anoopcs> hagarth, I am waiting for final round of review
12:32:49 <rjoseph> hagarth: This also have dependency on AFR lock healing. Also Sharding will not work correctly with mandatory locks
12:32:49 <hagarth> anoopcs: ok
12:33:44 <hagarth> rjoseph: not working with sharding should be acceptable as we have limited, specific use cases for sharding at the moment.
12:34:05 <rjoseph> hagarth: yes, we are treating that good to have
12:34:29 <hagarth> rjoseph, itisravi: are you considering lock + lease healing as a single effort? or would they need separate implementations?
12:34:46 <rjoseph> its a single effort
12:35:20 <hagarth> rjoseph: ok, thanks. let us determine how we can progress on that one after this meeting.
12:35:39 <rjoseph> hagarth: sure
12:35:49 <hagarth> anything else on mandatory locks?
12:36:15 <anoopcs> hagarth, Nothing other than reviews asap.
12:36:16 <hagarth> ok, let us move on to  Geo-replication/Glusterfind Enhancements
12:36:24 <hagarth> anoopcs: sure
12:36:49 <hagarth> aravindavk, kotreshhr: any updates on geo-replication/glusterfind ?
12:37:50 <hagarth> i haven't heard from them earlier, will check with them after this meeting.
12:38:28 <hagarth> moving on to tiering
12:38:33 <hagarth> dlambrig_: are you around?
12:39:08 <hagarth> guess not, moving on
12:39:09 <kotreshhr> hagarth: sorry for the delayed response. geo-rep/sharding patches are merged upstream!
12:39:20 <kotreshhr> hagarth: and 3.7 as well.
12:39:29 <hagarth> kotreshhr: fantastic!
12:39:41 <ndevos> rafi1: tiering update?
12:40:23 <kotreshhr> hagarth: And also tiering aware geo-rep patches are merged in master branch
12:40:57 <hagarth> kotreshhr: great
12:41:04 <kotreshhr> hagarth: not yet started with directory filters for geo-rep.
12:41:10 <hagarth> kotreshhr: ok
12:41:42 <hagarth> kotreshhr: any update on bareos  - glusterfind integration?
12:42:32 <ndevos> I think the Bareos people mentions they wont spend more time on integrating with glusterfind until the API/format is stable
12:42:50 <kotreshhr> hagarth: millind is actively working on it. A new enhancement which to add query option is merged in master recently. But need to check with him on about the integration.
12:43:29 <hagarth> ndevos, kotreshhr: ok
12:43:47 <hagarth> ok, let us move on to Kerberos for Gluster protocols
12:44:03 <hagarth> ndevos: any updates here?
12:45:59 <hagarth> let us hear from ndevos a bit later on this
12:46:13 <hagarth> next feature - 9.  AFR self-heal performance improvements:
12:46:26 <dlambrig_> hargarth- here
12:46:57 <hagarth> dlambrig_: we were wondering about the progress on volume expansion/subtraction + samba support with tiering
12:47:58 <dlambrig_> discussed with Ira how Samba configuration with tiering, we will start on that after the current release ships
12:47:58 * ndevos got distracted...
12:48:24 <hagarth> dlambrig_: are we still aiming to get samba support in 3.8?
12:48:24 <dlambrig_> volume expansion same, will get to that after current release
12:49:12 <ndevos> Kerberos: design is in the glusterfs-specs repository, more comments, improvement welcome
12:49:32 <ndevos> Kerberos: also, we could use some volunteers to work on some tasks :)
12:50:02 <hagarth> dlambrig_: ok, can I call out both these features as pre-design?
12:50:04 <dlambrig_> tiering could use some volunteers as well. step up , everyone , form a line
12:50:36 <hagarth> maybe we should send a consolidated list of features/tasks that need help on -devel.
12:50:37 <dlambrig_> hagarth: sure
12:50:46 <hagarth> dlambrig_: thank you!
12:51:42 <hagarth> ndevos: for kerberos, would the status be "Design ready, help needed for implementation" ?
12:51:56 <ndevos> hagarth: yes, that matches it
12:52:36 <hagarth> ndevos: thank you, have updated the etherpad
12:52:42 <hagarth> coming back to afr
12:52:42 <ndevos> hagarth: we have two Kerberos experts in Red Hat that reviewed the design, so that part looks ok :)
12:52:58 <hagarth> itisravi, kdhananjay: any updates here on performance improvements?
12:53:26 <itisravi> hagarth: a set of patches have been posted for entry sellfheal performance.
12:53:40 <itisravi> some work is still pending.
12:53:48 <hagarth> itisravi: ok
12:54:11 <itisravi> We are not targetting all the items listed in the md file for 3.8
12:54:32 <itisravi> i.e not everything in http://review.gluster.org/#/c/12257/1/in_progress/afr-self-heal-improvements.md
12:54:55 <itisravi> BUT, we are also looking at including richard's patches into 3.8
12:55:06 <hagarth> itisravi: what features are not being targeted to 3.8 from here?
12:55:16 <itisravi> hagarth: granular data selfheal
12:55:35 <hagarth> itisravi: ok
12:55:55 <kdhananjay> hagarth: There are performance improvements on the IO side too. Compound xlator for improving IO performance, which pranithk is to develop --  he seemed confident about being able to pull it off for 3.8.
12:56:13 <itisravi> right
12:56:21 <hagarth> kdhananjay: sounds very good, compound xlator might benefit kerberos too
12:56:44 <kdhananjay> and apparently DHT as well towards improving IO perf.
12:56:49 <hagarth> let us try to enumerate the afr improvements better in the feature tracking page for 3.8
12:56:52 <ndevos> "might benefit" not really, it is a dependency :)
12:57:13 <itisravi> I also want to see if we can do something on the throttling front.
12:57:18 <hagarth> ndevos: without knowing the actual details, I hesitated to make a strong statement there :)
12:57:44 <itisravi> Like using bit rot's token bucket filter as a server side xlator to regulate shd fops.
12:57:45 <hagarth> itisravi, kdhananjay: let us update this page with the specific enhancements targeted - https://www.gluster.org/community/roadmap/3.8/
12:57:55 <kdhananjay> hagarth: Sure.
12:58:03 <ndevos> hagarth: yeah, sure, but we'll need it for general Kerberos use-cases :)
12:58:35 <hagarth> itisravi: sounds very good, looks like there is a lot of work happening on afr.
12:58:48 * rjoseph need to jump to another meeting
12:58:53 <hagarth> moving on
12:58:56 <hagarth> Sub-directory export for native protocol
12:58:58 <itisravi> hagarth: only strech goals :)
12:59:07 <hagarth> itisravi: we will make some of them real :)
12:59:16 <hagarth> rjoseph: thank you for the updates!
12:59:23 <itisravi> it might be later than sooner, but definitely in  at some point in time:)
12:59:50 <hagarth> sub-directory export fails a few regression tests, I am looking into them. fairly confident that we will get this into 3.8 unless we stumble upon something dramatic.
13:00:03 <hagarth> i will be updating a feature page subsequently for this
13:00:23 <hagarth> moving on to DHT improvements
13:00:33 <hagarth> sakshi, susant: any updates here?
13:01:14 <sakshi> hagarth, for 'lookup-optimize on by default' the design is ready and in development
13:01:14 <susant> sakshi is working on making "lookup-optimize " default
13:01:45 <hagarth> sakshi: that would be great to have! is the design posted in gluster-specs or on -devel mailing list?
13:01:48 <susant> The lock migration design is ready according to Du and Pranith
13:01:58 <hagarth> susant: nice
13:02:09 <susant> Should be able make it to 3.8
13:02:16 <susant> *to make
13:02:37 <sakshi> hagarth, design not exactly posted, just discussed and viewd by dht team
13:03:24 <hagarth> sakshi: let us get it out onto gluster-specs
13:03:26 <ndevos> sakshi: we really need to make sure the designs get documented and posted in the glusterfs-specs repo :)
13:03:44 <hagarth> ndevos: please find a better name for glusterfs-specs soon ;)
13:03:56 <ndevos> oh, thats on me?
13:03:58 <sakshi> hagarth, will do
13:04:01 <hagarth> susant, sakshi: any updates on readdirp performance improvements?
13:04:01 * ndevos thinks
13:04:44 <susant> That is still under discussion
13:04:47 <hagarth> susant: ok
13:05:01 <hagarth> anything else on dht?
13:05:15 <susant> That's all
13:05:24 <hagarth> susant, sakshi: thank you!
13:05:27 <hagarth> moving on to  RichACL support for GlusterFS
13:05:44 <hagarth> rjoseph just moved out, will check with him later
13:05:57 <hagarth> moving on to 14 Share modes / Share reservations
13:06:12 <hagarth> poornimag: would you be able to provide an update on this one?
13:06:52 <rjoseph> Just a quick update: I sent the initial patch to github which will check and enforce richacl provided backend filesystem has richacl support
13:07:09 <rjoseph> not able to spend much time on this
13:07:22 <hagarth> rjoseph: ok, will sync up with you on this one later
13:07:34 <rjoseph> Now need to work on richacl enforcement within gluster
13:07:37 <rjoseph> sure
13:07:41 <hagarth> moving on to Inotify
13:07:43 <skoduri> hagarth, AFAIK rastar still has to put up the feature page/ design doc..but has been almost ready with core changes required for review
13:07:44 <rjoseph> me and poornima in the meeting
13:07:55 <rjoseph> *another meeting
13:07:57 <hagarth> skoduri: thank you for the update
13:07:59 <skoduri> hagarth, for share reservations ^^^
13:08:17 <hagarth> skoduri: got it
13:08:37 <hagarth> skoduri: I will also check with rastar once he is back
13:08:46 <skoduri> hagarth, sure
13:08:53 <hagarth> skoduri: we will defer inotify support for beyond 3.8, right?
13:08:58 <skoduri> hagarth, right
13:09:17 <hagarth> skoduri: thank you, have updated the etherpad
13:09:29 <hagarth> moving on to  trash translator improvements
13:09:43 <hagarth> jiffin: any updates? I think some patches are ready ?
13:10:04 <jiffin> hagarth: yes patch on trash xlators posted
13:10:08 <jiffin> upstream
13:10:19 <jiffin> https://review.gluster.org/#/q/message:%22features/trash%22+AND+age:1d+status:ope
13:10:23 <hagarth> jiffin: thank you, let us review those and get them in.
13:10:46 <jiffin> hagarth: I just started working on trash-helper xlator on client side
13:10:53 <hagarth> jiffin: ok
13:11:37 <hagarth> moving on to pNFS Layout Recall
13:11:49 <hagarth> skoduri, jiffin: any updates here?
13:12:06 <skoduri> we have put up feature page for review -http://review.gluster.org/#/c/12367
13:12:18 <skoduri> getting it reviewed from nfs experts first..
13:12:31 <skoduri> if we do not hear any major concerns shall circulate it to gluster-devel
13:12:41 <hagarth> skoduri: ok, sounds good
13:13:05 <skoduri> since its leverages on leases support, with core changes up for review we shall start working on layouts as well
13:13:18 <hagarth> skoduri: ok
13:13:47 <hagarth> well that completes the status update on the short list of proposed features for 3.8 :)
13:14:13 <hagarth> moving on to next topic
13:14:17 <hagarth> #topic Schedule for 3.8
13:14:18 <ndevos> + SElinux support is coming ;-)
13:14:29 <hagarth> ndevos: I know that there are a few more ;)
13:14:41 <hagarth> i did see xavih's big patch on ec too :)
13:15:03 <hagarth> me and raghu have made some progress on iscsi/block support as well
13:15:10 <hagarth> so let us see what all we end up with!
13:15:17 <hagarth> coming back to schedule
13:15:36 <hagarth> I am thinking of having a feature freeze in mid - late January
13:15:42 <hagarth> does that seem acceptable to all?
13:16:18 <hagarth> yes/no/maybe ?
13:16:27 <skoduri> maybe :)
13:17:01 <jiffin> for trash improvements it is K , since layout recall has dependency on lease I am not sure
13:17:40 <hagarth> right, if we miss a feature or two in 3.8.0 GA - we can always look at having them out in a subsequent .x release
13:17:41 <ndevos> hagarth: iscsi/block is building on the work that dlambrig_ did on iscsi-tgtd (or however its called)?
13:18:01 <hagarth> ndevos: there are other options too now with LIO, tcmu etc.
13:18:23 <hagarth> we have been playing around with lio tgtd backend
13:18:25 <jiffin> and for kererbos , if didn't get enough volunteers , "no" for kerberos ndevos??
13:18:25 <ndevos> hagarth: yes, hence my question :)
13:18:48 <ndevos> jiffin: we're working in the 'volunteer' part :)
13:19:11 <hagarth> ndevos: we intend providing an update on the iscsi bit soon in -devel
13:20:02 <ndevos> hagarth: I really would appreciate the "upstream first" thought a lot, not only code, also ideas and designs :)
13:20:10 <hagarth> ok, I don't see major objections for the schedule. tentatively let us work with mid - late Jan as feature freeze. we can arrive at other milestones in a subsequent meeting when we have more clarity.
13:20:22 <jiffin> ndevos: sooner the better :)
13:20:55 <ndevos> schedule looks good to me
13:20:56 <hagarth> ndevos: of course, just validating a few ideas with PoCs before hitting the MLs
13:20:59 <skoduri> hagarth, +1
13:21:35 <hagarth> I will aim to have one more status update meeting in January and we can take it from there.
13:21:57 <hagarth> having said that, I will also push out the last topic for today to that meeting :)
13:22:28 <hagarth> any last questions/thoughts/comments?
13:22:41 <ndevos> I guess we need more time for the next meeting?
13:23:01 <hagarth> ndevos: yes, maybe we can be better prepared before the meeting too
13:23:46 <hagarth> I was actually hoping to wrap this meeting in 45 minutes ;)
13:24:14 <ndevos> probably yes, send the announcement of the meeting, and the things that each feature owner needs to prepare
13:24:26 <jiffin> hagarth: we just took only 2x45
13:24:28 <hagarth> alright then, thank you all for being here and providing updates on 3.8. I will summarize our discussion here and send out a note on -devel.
13:24:37 <hagarth> jiffin: 6 short of 2 x 45 ;)
13:24:37 <jiffin> hagarth: thanks
13:24:51 <hagarth> #end-meeting
13:24:53 <jiffin> hagarth: right
13:24:55 <hagarth> #endmeeting