15:04:24 <JoeJulian> #startmeeting Gluster community weekly meeting
15:04:24 <zodbot> Meeting started Wed Jul 18 15:04:24 2018 UTC.
15:04:24 <zodbot> This meeting is logged and archived in a public location.
15:04:24 <zodbot> The chair is JoeJulian. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:04:24 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:04:24 <zodbot> The meeting name has been set to 'gluster_community_weekly_meeting'
15:04:50 * amye is here
15:04:59 * loadtheacc is here
15:05:45 <JoeJulian> There are no topics on the board...
15:05:48 * sankarshan is perhaps here
15:06:10 <sankarshan> (I had a question to put up when other topics are exhausted)
15:06:31 <amye> I can add a topic: any feedback on the release changes? (from x.y.z to just x)
15:07:09 <JoeJulian> #topic feedback on the release changes
15:07:44 <JoeJulian> I know that had been talked about a long time ago. Is that being done?
15:07:57 <amye> Yes, Shyam mailed the lists to say 'this is happening'.
15:08:03 <amye> Instead of 4.2, we're just moving to 5.
15:08:49 <amye> My hope is that it'll make understanding what the current version is a lot simpler. Also we won't end up with 'the 13th release in this cycle'
15:09:10 <JoeJulian> Well that's certainly going to be interesting to see how that pans out in #gluster.
15:09:40 <JoeJulian> So they're all stable versions from now on...
15:09:43 <amye> Should we do more notes about it?
15:09:44 <amye> Yes.
15:10:46 <amye> That was all I had, we can move towards other topics if release numbering is crickets.
15:10:49 <JoeJulian> Interesting - especially when a brand new feature is released with no production history.
15:11:37 <JoeJulian> Perhaps there should be some sort of feature gating (like kubernetes does) that doesn't enable beta features unless specifically enabled.
15:11:53 <nigelb> I believe that's already the plan.
15:11:58 <amye> That's a good thought, I will check Shyam's email to see if we've specifically called out features like that.
15:12:17 <amye> https://lists.gluster.org/pipermail/announce/2018-July/000103.html
15:12:22 <JoeJulian> cool. I'm sorry I missed that email. So many...
15:12:58 <amye> Doesn't look like we have 'tech preview'  or anything called out in that mail, but that was the plan.
15:13:06 <amye> Glad I brought it up then.
15:13:58 <JoeJulian> Ok, if there are no other topics...
15:14:16 <amye> sankarshan, did you have anything?
15:14:32 <sankarshan> amye oh yes! I did. should i go ahead?
15:14:39 <amye> Absolutely.
15:15:18 <sankarshan> thank you. so, i wanted to bring up the notion of seeking feedback from users as to whether they see any specific patterns emerging in terms of issues at releases
15:15:37 <sankarshan> For instance, if they use GeoReplication heavily, are we seeing more issues or, some concerning things and such
15:16:00 <JoeJulian> #topic seeking feedback regarding patterns of issues from releases
15:16:35 <amye> We already do the release retrospectives, but that's focused on the actual release process (kindof.)
15:16:39 <JoeJulian> The biggest issue I've had asked in #gluster regarding geo-rep at all is "how can I find out why it's failed".
15:17:40 <sankarshan> ok . and has there been much progress made?
15:17:41 <JoeJulian> I wonder if ML could be applied to #gluster logs to find any sort of trends.
15:17:54 <amye> Does the bot do anything like 'count how often this term'.. You read my mind.
15:18:11 <JoeJulian> It doesn't. It would require something a bit more AI.
15:18:17 <sankarshan> I was wondering how useful would be to have an ELK based stack to chew logs for patterns
15:18:32 <amye> That could be a way to solve it.
15:18:40 <JoeJulian> That would be cool if we could get enough donated logs.
15:18:57 <JoeJulian> The problem with logs is when a problem isn't in it.
15:19:05 <sankarshan> there is a problem there in sharing of logs though. but yes.
15:19:34 <JoeJulian> Like "heals are way too slow in 3.13 so I had to roll back to 3.12"
15:19:38 <sankarshan> the other part is - how to crowd source trouble shooting processes followed and perhaps make them part of a reviewed set of documentation on gluster.org
15:20:48 <JoeJulian> I've been asked to do that so many times, but the problem I have is that I don't know how to diagnose things until people ask the question.
15:20:49 <amye> Ask on the mailing lists? I'm trying to think of where else we could ask.
15:21:32 <sankarshan> (in a general sense of this conversation - I am trying to find ways to understand how our releases are doing in terms of specific capabilities as well how we can improve on our troubleshooting section)
15:21:42 <JoeJulian> If there was a clear diagnostic path for a particular problem, that problem should have been coded around.
15:22:06 <sankarshan> that i agree. a problem that is today and being diagnosed, should be addressed as a fix in future
15:22:10 <amye> You want the months long view of releases and problems rather than the first few weeks for a new release.
15:22:30 <sankarshan> i do. i also want to understand how to outreach this more to get feedback
15:22:36 <JoeJulian> You would need that. Upgrading in production is a months long process after a release.
15:22:39 <sankarshan> for any project feedback is always a bit poor anyway
15:24:01 <JoeJulian> I hate to say it, but the only way to get most people to even think about giving feedback is if you can somehow nag them at the cli when a version change is detected.
15:24:02 <sankarshan> anyway - the idea is a bit nebulous right now, so i guess i haven't framed it well. i'll soak in this a bit more before i post to lists
15:24:40 <amye> I like the idea of having an ELK stack chew on the #gluster logs for patterns, we could also do a specific email on the mailing lists (like you're asking), I can do a roundtable topic on this at mountpoint, lots of options.
15:25:29 <sankarshan> (that's all I had to contribute, if it was not obvious)
15:26:35 <JoeJulian> I wonder if it would be considered evil if a release checker was added to glusterd to log upgrade availability. That call-in would be a metric.
15:26:36 <amye> That's a good thing to think about. How can we get more feedback from upgrades, not just releases.
15:27:29 <JoeJulian> #action Think about how to get more feedback from upgrades, not just releases
15:27:39 <amye> JoeJulian, slightly? Given that it doesn't already exist, introducing it now would require a lot of communication.
15:28:02 <JoeJulian> If it was disabled by default it might be ok.
15:28:26 <amye> That could work.
15:29:11 <JoeJulian> #topic hackmd.io release updates
15:29:26 <JoeJulian> They are all out of date. Can we have the release managers update that please?
15:29:51 <amye> I don't think we use those anymore, sadly.
15:30:00 <amye> Drop a note to maintainers@?
15:30:00 <JoeJulian> And can we also have heketi, ganesha, and anthill added by their maintainers?
15:30:27 <JoeJulian> #action drop a note to maintainers
15:31:10 <amye> Thank you.
15:31:16 <JoeJulian> kkeithley: Could you find out about whether someone would be willing to add that from ganesha?
15:31:54 <JoeJulian> Ok, gotta run. Thanks.
15:32:12 <JoeJulian> #endmeeting