Tuesday, 2017-12-12

*** tdasilva has joined #openstack-tc00:05
*** mwhahaha has joined #openstack-tc00:05
*** fdegir has joined #openstack-tc00:10
*** DuncanT has joined #openstack-tc00:10
*** diablo_rojo has quit IRC00:39
*** kumarmn has joined #openstack-tc01:11
*** liujiong has joined #openstack-tc01:44
*** kumarmn has quit IRC01:56
*** mriedem has quit IRC02:15
*** liujiong has quit IRC03:02
*** liujiong has joined #openstack-tc03:06
*** ianychoi has joined #openstack-tc03:08
*** kumarmn has joined #openstack-tc03:13
*** kumarmn has quit IRC03:24
*** kumarmn has joined #openstack-tc03:24
*** kumarmn has quit IRC03:29
*** kumarmn has joined #openstack-tc03:29
*** kumarmn has quit IRC03:53
*** kumarmn has joined #openstack-tc04:20
openstackgerritDai Dang Van proposed openstack/governance master: Update policy goal for watcher  https://review.openstack.org/52729904:22
*** kumarmn has quit IRC04:23
*** robcresswell has quit IRC05:39
*** fungi has quit IRC05:45
*** fungi has joined #openstack-tc05:48
*** kumarmn has joined #openstack-tc06:24
*** kumarmn has quit IRC06:28
*** dtantsur|afk is now known as dtantsur06:52
*** robcresswell has joined #openstack-tc07:39
-openstackstatus- NOTICE: Our CI system Zuul is currently not accessible. Wait with approving changes and rechecks until it's back online. Currently waiting for an admin to investigate.08:49
*** cdent has joined #openstack-tc08:55
cdenttc-members and everyone interested: it's office hours time09:00
cdentflaper87: are you responsible for this terrible cold I have?09:01
flaper87I'm around in case there are topics to discuss. I don't have anything to bring up.09:01
flaper87cdent if you caught it surfing, then yes09:02
cdentno, it's meant no surfing this week :(09:03
-openstackstatus- NOTICE: Zuul is back online, looks like a temporary network problem.09:09
cdentLooks like the only review that is stuck or contentious is still the interop tests one: https://review.openstack.org/#/c/521602/09:09
flaper87yeah, I caught up with the discussion yday but I haven't replied yet09:13
*** jpich has joined #openstack-tc09:13
flaper87it is indeed contentious and full of trade-offs09:14
ttxsorry for lateness, had dentist appointment09:32
ttxcdent, flaper87: I can summarize the 4 k8s/openstack discussions we had if you are interested09:34
flaper87ttx: yes09:34
ttxSo the first one was on open dev tooling09:35
ttxK8s community is struggling a lot with GitHub, apparently a lot of the discussions in their dev "invitation-only" day on tuesday were around that09:35
ttxMost declared bankruptcy on using GitHub issues09:36
ttxthere were lots of parallels with OpenStack there09:37
ttxThey were tempted to explore feature branches more09:37
ttxWe told them to be careful/conservative around those09:37
ttxThey are considering moving to gitlab, actually09:38
ttxalthough Gerrit is still around since Google people love it09:38
ttxThings that they manage to do with their tools that would be nice for us to be able to do: core review delegation09:39
flaper87oh mmh, self-hosted gitlab?09:39
ttxflaper87: unclear09:39
flaper87what's core review delegation?09:39
flaper87sounds like something I have proposed in the past09:39
ttxA core reviewer delegating his core review rights on a specific patch to a non-core reviewer09:39
ttxyes we did discuss it09:40
ttxIt looks like tagging issues with SIG names is also pretty efficient09:40
flaper87yeah,I think I brought that up again at the forum in SYD but not many ppl seemed to buy-in09:41
flaper87perhaps it is worth discussing it again09:41
ttxanyway, no miracle solution there -- just a feeling that they were having a lot of the issues we were having09:41
flaper87hopefully our feedback was useful to them09:41
ttxThe second discussion was on open innovation dynamics09:41
ttxThat one was eye-opening to me09:42
ttxThey are struggling a lot with strategic contributions09:42
ttxmuch more than we do/did09:42
ttxwhich is not apparent from the hype and the press09:43
ttxIn our case we are struggling a bit now, but during ramp-up phase we had people covering everything09:43
ttxThey are struggling *during* ramp-up09:44
ttxAnyway, we brainstormed and came up with two ideas09:44
ttxCaleb Miles and myself will co-author a blogpost around the importance of strategic contributions in those kinds of projects09:45
ttxAnd we'll push the respective corporate sponsors of our respective foundations to report "how they contribute to the project"09:46
ttxwhich is a positive way to encourage them to do the rigth things09:46
ttxbeyond silly numbers09:46
ttxHOW/WHAT rather han how many09:47
flaper87or how much09:47
cdentthat's a good idea09:47
ttxright. We have employee X who owns startegic task Y. Or We organize event Z which benefits the whole local community09:48
cdentsomething a bit more narrative, instead of numbers09:48
ttxcdent: exactly09:48
ttxI can't take credit, idea is Caleb's09:48
ttx(who btw attended every discussion, and is awesome)09:48
ttxanyway, we'll try to roll that out on this year report. Might be tricky given limited time09:49
ttxAnother insight of that discussion was the feature creep situation, which they are struggling with09:49
cdentI suspect that even a limited report, hinting at bigger reports to come, would be useful09:49
ttxobviously rang a bell09:49
cdent(big bell)09:50
ttxWe arrived at the conclusion that most of it is systemic to the way we do things, but you can still try to put safeguards in place early on09:50
ttxLike affirm scope aggressively, and say what you WON'T do09:51
flaper87s/early on/any time/09:51
flaper87early is better but...09:51
cdenthaving clear "not that" boundaries would be great for everyone, especially in the adjacent community discussions: "we don't do that, group X over there does"09:52
ttxI thought they would have a better time with feature creep than we did due to K8s being more advanced when it was open09:52
flaper87Do we have any other actionable items for ourselves other than the blog post you'll be writing?09:53
ttxbut they are struggling from their startup ecosystem and general NIH syndrome due to hyped community09:53
ttxI'm still wrapping my head around it09:53
ttxDiscussion 3 was around governance09:53
ttxWe arrived at the conclusion that a lot of the misundertanding/critics there was coming from the fact that our systems don't align09:54
ttxLike the CNCF TOC having a completely different role than openstack TC09:54
flaper87oh, mmh, intersting09:55
ttxThey say we get in the way of project governance, we say they are apponted corporate shills09:55
ttxBut their TOC is just selecting independent projects09:55
ttxnot even trying to paint an overarching story09:55
cdentwe seem confused about that internally sometimes too :)09:56
ttxThe equivalent of the TC is the K8s steering committee09:56
flaper87yeah. Although, it's interesting how a similar group of people have a complete different purpose in both communities09:56
ttxan elected body to make final calls and paint the overarching story at the K*s level09:57
ttxthey just don't have an equivalent at CNCF-level, by design09:57
ttxTheir SIGs look more like our project teams, although the code ownership is a parallel structure09:57
ttxi.e. the only way to break review disagreement is actually to go to steering committee09:58
ttxThe most eye-opening for me is:09:58
ttxwe build a dual safety-valve system with PTL and TC, they have a single safety-valve system with steering committee09:58
ttxTheir SOGs can have a series of leads, they don't really elect them09:59
flaper87so, projects don't have a "technical lead"09:59
ttxWe concluded that there were good and bad sides to that09:59
ttxOn the good side, their SIG leads can share the leading work10:00
ttxOn the bad side, there is no good moment to step down10:00
ttxso, different styles of burnout but burnout all the same10:00
ttxAlthough that got me thinking a bit10:01
ttxLike do we really need PTL elections10:01
ttxMind you, they ask themselves the same questions, from the other perspective :)10:02
ttx"would we be better with regular SIG lead elections"10:02
cdentI thikn the PTL elections provide a visible boundary opportunity10:03
ttxwhich shows that those are complex issues, and that discussing them was a great idea :)10:03
flaper87there will be an election of some sort, regardless. Someone will volunteer, etc10:03
ttxDiscussion 4 was the least attended (but then also the most specialized one) -- release management10:03
ttxWe were a bit tired by that point, but covered things like stable branches, communication around features / release notes...10:04
ttxThey are struggling to get good information from devs, we shared reno and other tools we use10:05
ttxI probably forgot important topics. dhellmann took a lot of notes, he probably has more to share10:05
ttxbut that's what resonated with me10:06
cdentBy "good information" do you mean "what's in this release"?10:06
ttxcdent: not only. Separating what's information for developers (commit message) from what's information for users (release notes)_and what's product/marketing info (release highlights)10:08
* cdent nods10:08
ttxCurrently they are relying on commit messages for most10:08
ttxwe use 3 different systems10:08
ttxcommit messages / reno / release highlights10:08
ttxwhich I think help you switch style10:09
ttxalthough we still need to see how the latest will go10:09
ttxSo yeah, overall we all agreed it was beneficial to hold those discussions and that we should do it again10:10
ttxhopefully not hanging as much by the thread next time10:10
flaper87sorry had to jump on a call10:11
ttxThanks to dims and others who helped beating the drum of interest and getting k8s people there10:11
flaper87so, next time it's kubecon EU10:12
ttxon that first lunch I had no idea who would end up coming :)10:12
flaper87hahahaha, it sounds like it turned out to be useful10:12
ttxor OpenStack Summit Vancouver10:12
flaper87kubecon comes first, though. But it would be good to do it in both confs, I guess10:13
ttxkubecon eu is first10:13
ttxCFP is open. I might try to submit panel(s) to have those on the schedule10:16
flaper87ttx: need help? I'm planning to submit talks to kubecon too10:17
ttxI'll let you know. I'd rather be a panelist than a moderator for those, and ideally the mod would be someone with ties to both communities.10:18
ttxor someone seen as neutral10:18
flaper87yeah, I don't feel like moderating to be honest10:19
cdentAt what point does the "have fewer meetings to be more inclusive" extend to "rely less on confrences"? (I'm not disputing their value, especially in establishing relationships, but their prominence/importance can be excluding)10:22
flaper87cdent: they can be excluding if discussions and/or decisions depend on them.10:23
flaper87cdent: for example, it would be great if the conversation that started at kubecon NA would continue without needing a conference10:23
flaper87but I would argue that not doing PTG/Forum would be extremly damaging for the community and the project, despite the fact that they both are not "attendable" by everyone10:24
cdentyeah, I think there's a distinction between conference and PTG10:27
cdentI often think things like the PTG should happen more often10:28
*** liujiong has quit IRC10:29
cdentand that platinum and gold members should commit far more to paying for people to attend10:29
*** rosmaita has joined #openstack-tc13:04
*** kumarmn has joined #openstack-tc13:25
*** kumarmn has quit IRC13:29
openstackgerritMerged openstack/project-team-guide master: Removing the cross-project team  https://review.openstack.org/50631113:30
openstackgerritMerged openstack/governance master: Update shade team metainfo  https://review.openstack.org/52351913:36
openstackgerritMerged openstack/governance master: Rename shade team to OpenStackSDK  https://review.openstack.org/52352013:42
openstackgerritMerged openstack/governance master: Move os-client-config to OpenStackSDK from OpenStackClient team  https://review.openstack.org/52424913:42
openstackgerritMerged openstack/governance master: Update policy goal for mistral  https://review.openstack.org/52478213:59
openstackgerritMerged openstack/governance master: Add mistral-tempest-plugin to mistral project  https://review.openstack.org/52486813:59
openstackgerritMerged openstack/governance master: Add congress-tempest-plugin to congress project  https://review.openstack.org/52506613:59
openstackgerritMerged openstack/governance master: Update tempest plugin split goal for Tacker team  https://review.openstack.org/52548914:00
*** mriedem has joined #openstack-tc14:03
cdentrobcresswell, mugsie this https://etherpad.openstack.org/p/ptl-meeting kind of faded out. Is there anything we should consider doing to keep the needs it represents visible? It seemed like the idea of a meeting that obliged PTLs to do more (at least in a synchronous fashion) was not going to float, but the problem remains.14:11
openstackgerritMerged openstack/governance master: remove docs:follows-policy tag  https://review.openstack.org/52421714:11
fungion tagging proposed changes with sig names, next gerrit upgrade will get us "hashtags" we can use to decorate them with whatever arbitrary strings we like14:15
ttxfungi: nice!14:18
fungiwe could have stretched to get a sufficiently high gerrit version in our last upgrade, but it needed newer java, hence newer ubuntu, so either an in-place distro upgrade (which we usually avoid) or a new instance with a new ip address (which annoys third-party ci operators in corporate networks with draconian firewall admins)14:21
fungiand also that gerrit release was very much bleeding edge at the time14:21
funginow it's had time to simmer a little14:21
dhellmannduring the tooling discussion, we talked about various bots that they and other groups like python-dev use to manage github notifications14:22
dhellmannI also thought it was interesting that they tried "assigning" pull requests to reviewers based on owners files14:23
dhellmannI think they said they stopped doing that, since it wasn't really leading to more reviews14:23
dhellmannthey also mentioned that it can be difficult to run some of their test jobs locally and contributors are starting to rely on the automated CI for those, much like what has happened in our community14:24
dhellmannI thought it was interesting that for some of their integration test jobs, some projects only test against released kubernetes14:25
dhellmannthat means they have some stability in the jobs, until the next kubernetes release14:25
dhellmann(which happens roughly quarterly)14:25
dhellmannwe talked about testing kubernetes patches on top of openstack using zuul, and there was some interest but I can't say anyone in the room was excited enough to commit to doing the work right there14:26
dhellmannit also wasn't rejected out of hand, though14:26
dhellmannfrom the governance lunch, one of the early things that caught my attention was their "ladder" for new projects14:27
dhellmannthey have more formal inception, incubation, etc. stages14:27
dhellmanneven kubernetes is still technically incubated14:27
dhellmannI think anyone can propose a project for inception, but for incubation there have to be at least 2 companies contributing to it14:28
dhellmannwe compared that to our diverse affiliation tags14:28
dhellmannit achieves somewhat the same thing, but more formally translates into a measure of maturity14:28
dhellmann(their model does)14:28
dhellmannthe split between SIGs and code ownership was confusing to me. It seems a SIG is responsible for designing and making decisions, but the code is potentially owned by other people?14:29
cdentthat sees ripe for the same pwg problems we've seen in this part of the world?14:30
dhellmannthat came up directly in the discussion of the openstack-sig later in the week when angus pointed out that he had written a bunch of the code that the new sig was now driving, but he wasn't actually involved in creating the sig14:30
dhellmannin addition to the project ladder, they apparently have a more formal contributor ladder for helping to onboard people14:31
dhellmannI haven't looked for the actual documents for that, but maybe that's something the first-contact sig could look into14:31
dhellmannI know some of our teams have some guidelines, but I don't think they all do14:31
dhellmannduring their ToC public meeting (they also have a private meeting) one thing I made a note of is that they actively seek out new projects to "fill gaps"14:33
dhellmannthey didn't give any real details about what sorts of gaps they were trying to fill, though14:33
cdentthe interesting part of that to me is that that implies they are actively keeping track of gaps (within the ToC)14:33
dhellmannyeah, that was less clear. it was said as a "this is one of the things the toc does" bullet item and there was very little detail14:34
dhellmannit would be interesting to see what form that really takes14:34
dhellmannsomeone in that meeting suggested having more end-user involvement in the CNCF as a whole, so I think they're running into some of the same vendor-focus issues we've had in the past14:35
dhellmannand envoy was mentioned as an example of a project that was created by an end-user group rather than a vendor14:36
dhellmannduring the release session, Caleb mentioned that they support 3 versions of kubernetes at one time and support upgrade from N-2 to N as part of their policy14:36
dhellmannwith quarterly releases, I imagine it's a little easier to do that and still drop support for deprecated things relatively quickly14:37
dhellmannif my math is right, it seems like they're not supporting a given release for a full year14:37
mugsieand with so many features in "alpha" or "beta" there is less of a support / compatibility burden, right?14:38
dhellmannI did get the impression from some of the presentations that they are not plagued with vendors trying to differentiate through installation tools14:38
dhellmannmugsie : they did announce 1.0 stable status for several APIs last week, so that's likely to be changing14:38
mugsiedhellmann: is installers not the only differencation that k8s vendors have right now?14:39
dhellmannthe discussions were interesting, and I agree these meetings are probably good first meetings and I hope we are able to schedule some others14:39
-openstackstatus- NOTICE: We're currently seeing an elevated rate of timeouts in jobs and the zuulv3.openstack.org dashboard is intermittently unresponsive, please stand by while we troubleshoot the issues.14:40
mugsieI know it was our main one when I was working on CaaS14:40
dhellmannmugsie : my impression was they were trying to standardize somewhat, but I may have misunderstood something14:40
dhellmannmaybe there is only one "community" installer?14:40
mugsieyeah - there is a single community installer14:40
dhellmannthat's probably what I was hearing about then14:41
mugsiebut e.g. Suse had a Salt based one, RH has OpenShift(?), etc14:41
dhellmannbecause, yeah, I'm sure openshift's installer doesn't work the same way as a vanilla kubernetes installer14:41
dhellmannthey're also working on API-driven compliance testing, much like what we have14:41
mugsiethere was some good slides I saw about k8s distros vs a community distro on twitter14:42
dhellmannif you can find that link, I'd be interesting in looking at it14:42
dhellmannI think they said it might be a week or more before the videos go up14:43
*** kumarmn has joined #openstack-tc14:43
dhellmannSarah Novotny's closing keynote from friday was really quite good14:44
mugsiedhellmann: https://schd.ws/hosted_files/kccncna17/ac/KubeCon_2017_-_Kernels_and_Distros.pdf14:48
*** marst has joined #openstack-tc14:50
ttxre: CNCF project ladder, I think two organizations need to be vouching for it, not really contribute to it14:51
ttx(for it to enter incubation)14:51
ttxMost of their inclubated projects (with the notable exception of Kubernetes) woudl actually be considered single-vemdor with our metrics14:52
*** mriedem has quit IRC14:58
mugsiettx: would k8s reach our definition of diverse?14:59
ttxmugsie: no14:59
smcginnis+1 to both continuing those cross-community discussions at kubecon-eu and for trying to get a panel discussion.15:00
smcginnisttx: Maybe dims would be a good moderator for that?15:00
ttxhttps://devstats.k8s.io/dashboard/db/companies-stats?orgId=1 -- getting better though15:00
ttxGoogle+Redhat still above 50% for most metrics15:01
ttxAlso worth noting they are past peak activity15:01
smcginnisI would expect that to climb again though.15:06
ttxyes -- was just wondering why it dropped since Sept15:07
smcginnisMaybe lead up to release and kubecon? Ours usually ramps up then, but maybe they are better about locking things up earlier.15:07
ttxhmm probably a Google internal thing15:08
smcginnisBut as more vendors jump in, I would bet that spikes again.15:08
*** mriedem has joined #openstack-tc15:09
openstackgerritJeremy Stanley proposed openstack/governance master: Add an openstack/self-healing-sig repository  https://review.openstack.org/52532815:22
*** diablo_rojo has joined #openstack-tc15:57
robcresswellcdent: re: earlier ping, yeah, it seems that way. To be honest, my time on OpenStack has pretty much come to an end in the past couple of months, so I'm not around to drive it further :/16:08
robcresswellIt'd be nice if it was captured as a side note in any future docs, but I've little to add to it at the moment.16:08
cdenta) bummer for us but hopefully good for you, b) no worries on the driving the issues will remaining, driving at their own pace16:09
*** dtantsur is now known as dtantsur|afk16:28
fungipriority follows pain16:47
dimssmcginnis : i haven't thought about kubecon-eu yet16:52
pabelangerdhellmann: re: commit to doing work on zuul, personally, I think it would be better to show up with working zuul, running soe of their tests. There is a large overhead to admin / ops of zuul, this is the most exciting part of the github connection driver we have today.16:58
dhellmannpabelanger : yeah, I meant doing the work of integrating our zuul instance with their repos to run tests on their patches (and ours I guess)17:40
*** david-lyle has joined #openstack-tc17:40
pabelangerdhellmann: I did talk with dims at summit about how we could maybe help bootstrap. We have some good discussions about adding some k8s bits into devstack, just haven't had time yet to work on it17:41
dhellmannpabelanger : yep17:45
cdentI wish there was more of me, because I'd love to be able to devote some time to learn up on zull17:46
* smcginnis knows the feeling17:47
* dhellmann pictures zull as the "good" twin of zuul17:47
cdentI'd also like to learn up on the openstack thing too.17:50
cdent>3 years of working on it, still got no idea17:50
*** jpich has quit IRC17:53
fungiyeah, i probably have a lot deeper insight into how openstack (the software) works than i think i do, but i really don't feel like i have much of a handle on any of it18:17
fungispent so much of my time focused on automation above the api layer18:17
fungirather than actually running openstack itself (the infra-cloud deployment and occasional troubleshooting of devstack-based ci jobs aside)18:18
fungifrom the vmt side of things, i'm constantly having to remind myself to avoid jumping to conclusions about how things are implemented under the covers18:19
fungior how production deployments are using certain features18:19
fungibecause i rarely guess correctly in those cases18:20
cdentIt's a weird feeling isn't it?18:21
fungiit's like i exist in the demilitarized zone between the operators, users and developers18:21
cdentI think there are _plenty_ of developers (like me in this case) who don't have much of a clue about what's going on beyond their particular speciality18:22
cdentAnd while I think that's normal, and mostly okay, it's ... disconcerting (for me)18:23
fungii suppose it's far too much detail for any mortal to grasp18:24
fungiif only vish were here ;)18:24
*** cdent has quit IRC19:20
*** cdent has joined #openstack-tc19:36
*** tonyb_ has quit IRC19:52
*** tonyb has joined #openstack-tc19:52
*** cdent has quit IRC20:02
-openstackstatus- NOTICE: The zuul scheduler has been restarted after lengthy troubleshooting for a memory consumption issue; earlier changes have been reenqueued but if you notice jobs not running for a new or approved change you may want to leave a recheck comment or a new approval vote20:16
*** openstackgerrit has quit IRC20:34
*** alex_xu has quit IRC20:34
*** notmyname has quit IRC20:34
*** purplerbot has quit IRC20:35
*** amrith has quit IRC20:35
*** alex_xu has joined #openstack-tc20:39
*** notmyname has joined #openstack-tc20:41
*** amrith has joined #openstack-tc20:43
*** openstackgerrit has joined #openstack-tc20:43
openstackgerritKendall Nelson proposed openstack/governance master: Add Storyboard Migration to Rocky  https://review.openstack.org/51387520:43
*** kmalloc has quit IRC20:45
*** kmalloc has joined #openstack-tc20:45
*** kumarmn has quit IRC21:01
*** dklyle has joined #openstack-tc21:48
*** david-lyle has quit IRC21:48
*** flwang has quit IRC22:06
*** flwang has joined #openstack-tc22:19
*** marst has quit IRC23:08
*** chandankumar has quit IRC23:22
*** chandankumar has joined #openstack-tc23:23
*** chandankumar has quit IRC23:29
*** chandankumar has joined #openstack-tc23:34

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!