Thursday, 2018-02-15

*** kumarmn has joined #openstack-tc00:03
*** kumarmn has quit IRC00:07
*** asdffrfr has joined #openstack-tc00:17
*** asdffrfr has quit IRC00:27
*** mriedem has quit IRC00:50
*** openstackgerrit has joined #openstack-tc01:12
openstackgerritJeremy Stanley proposed openstack/governance master: Update Rocky PTLs
fungidiablo_rojo: persia: ^01:13
diablo_rojoDone :)01:17
*** kumarmn has joined #openstack-tc01:35
*** kumarmn has quit IRC01:41
*** harlowja has quit IRC03:08
*** kumarmn has joined #openstack-tc03:30
*** ykarel|away has joined #openstack-tc03:34
*** kumarmn has quit IRC03:34
*** rosmaita has quit IRC04:00
*** harlowja has joined #openstack-tc05:15
*** coolsvap has joined #openstack-tc05:18
*** harlowja has quit IRC06:14
*** eumel8 has joined #openstack-tc08:03
*** ykarel|away is now known as ykarel|lunch08:04
*** ykarel|lunch is now known as ykarel08:40
*** kumarmn has joined #openstack-tc08:42
*** kumarmn has quit IRC08:46
*** jpich has joined #openstack-tc08:59
*** ykarel_ has joined #openstack-tc09:06
*** ykarel has quit IRC09:08
*** coolsvap has quit IRC09:08
*** ykarel_ has quit IRC09:15
*** dtantsur|afk is now known as dtantsur09:21
*** ykarel has joined #openstack-tc09:22
*** cdent has joined #openstack-tc09:45
*** ykarel is now known as ykarel|away09:58
*** coolsvap has joined #openstack-tc10:07
*** ykarel|away has quit IRC10:08
*** ykarel has joined #openstack-tc11:04
*** ykarel has quit IRC11:22
*** pbourke has quit IRC11:30
*** pbourke has joined #openstack-tc11:32
*** coolsvap has quit IRC12:36
*** kumarmn has joined #openstack-tc12:42
*** kumarmn has quit IRC12:47
*** rosmaita has joined #openstack-tc13:47
*** mriedem has joined #openstack-tc14:01
*** dtantsur is now known as dtantsur|brb14:02
*** kumarmn has joined #openstack-tc14:04
*** dtruong has quit IRC14:40
*** dtruong has joined #openstack-tc14:41
*** dtantsur|brb is now known as dtantsur14:49
*** ykarel has joined #openstack-tc14:50
ttxoffice hour?15:02
* cdent was distracted by giving mordred a sec15:02
ttxDidn't have much, about to leave for a week vacation snowboarding in the French Alps15:02
cdentBit early for you EmilienM ?15:02
EmilienM7am? lol15:03
EmilienMI almost did half of my working day :D15:03
cmurphyttx: your life sounds awful15:03
ttxsmcginnis, pabelanger: you're on point to fill the "Rocky" section of the Monday PTG presentation, i won't be much around to send reminders15:03
smcginnisttx: Yep, got it.15:04
ttxcmurphy: I'll send pictures.15:04
smcginnispabelanger: We should discuss at some point. I have the presentation up and was planning on adding a few bullets at some point.15:04
ttxA couple more +1s on and I'll approve it tomorrow15:05
cdentis every tc member going to be at the ptg?15:05
ttx(Rocky PTLs)15:05
ttx(before I leave)15:06
ttxcdent: good question, I think so15:07
smcginniscdent: I will be there.15:07
fungias will i15:07
dhellmannI will be15:07
ttxdims, johnthetubaguy : you'll be there ?15:07
cmurphyptg roll call15:07
ttxall the others I think I alerady saw on the list15:07
dimsttx : yep. will be there15:08
cdentI added this topic to as well, but I was recently asked internal to my employer what being a "base service" really means in terms of usefulness/usability? The particular concern is with etcd, which only have partial support in tooz for group coordination15:08
johnthetubaguyI should be at the ptg, I am registered, etc15:08
dhellmanncdent : the idea is that if an openstack project wants to rely on one of those services being present to implement a feature, that's OK.15:09
ttxyes "can assume to be present in an openstack cloud"15:09
ttxsince there is a bit of chicken and egg there15:10
cdentdhellmann, ttx: yeah, so in this case is it "simply" a matter that tooz hasn't caught up?15:10
dhellmannit's not clear whether we need tooz to have full support in some way, since it's supposed to be an abstraction layer around similar services, right?15:10
cdentdhellmann: the missing functionality is (need to look for link) ...15:10
dhellmannI think once we added etcd to that base layer list, we expect projects to be able to talk to it directly15:11
dhellmannusing it's client15:11
dhellmanninstead of a wrapper15:11
cdentIs that last bit true. I thought we had declared tooz to be the way to go?15:12
dhellmannI don't remember15:12
smcginnisMy understanding is the same as dhellmann. Until we accepted it as a base service, tooz was just a possible abstraction to it. Now code can assume and use etcd directly.15:12
dhellmannif tooz provides integration with oslo.config, it makes sense to use it15:12
mugsieetcd + eventlet are not friends unfortunately15:12
smcginnisdims: Weren't you the one that drove the etcd acceptance?15:12
mugsieso for our usage of tooz, we can't use etcd15:12
dimssmcginnis : yep15:13
dhellmannso anyone who wants to use tooz with a feature of etcd not yet supported in tooz could work on those patches15:13
mugsiedhellmann: etcd3 interface + eventlet just don't work together, eventlet will hang forever if you try to connect15:14
smcginnisdims: Was that the reason for etcdgw? ^15:14
dhellmannyet another reason to get rid of eventlet15:14
dimsmugsie : i am maintaining this layer -
mugsiedims: that didn't do grouping did it? I thought the gateway couldn't do the active push required for member join / leave notifications15:15
dimsmugsie : if we need it, we should be able to add it15:15
* fungi wonders exactly how many reasons we need for getting rid of eventlet before a sufficient force force of developers and reviewers appears to actually make it happen15:16
mugsiefungi: I would *love* to remove it15:16
cdentmugsie: I'm drawing at brief memories, but yeah, what smcginnis  said15:16
smcginnisLast minute rocky goal? Get rid of eventlet? :D15:17
mugsieit is just so ingrained at this point, I think it would take multiple cycles15:17
cdent(sorry that message started before the phone rang and I hit return after getting off the phone, so it's a bit out of sync)15:17
* smcginnis could just imagine the pushback on that proposal.15:17
dimsLOLOL, need a magic wand15:17
mugsiesmcginnis: lets go :D15:17
cdentI think would should at least make sure that new things don't put it in15:17
cdentbut so many services copy from the existing services that it happens anyway15:18
pabelangersmcginnis: ttx: ++15:18
mugsiecdent: ++15:18
pabelangersmcginnis: yah, same. I think each adding a few points first, then syncing to review is a good first step15:18
mugsiebut oslo.service uses it as a base unfortunately so people starting with that will just have it15:18
smcginnispabelanger: ++ sounds like a good plan to me15:18
* fungi just now looked at
fungioff to a bad start there i guess15:18
fungibrand new project which is likely to try for official status during rocky15:19
smcginnisI've heard "eventlet is bad" for quite a while now, but I don't think I've seen a better recommendation.15:19
smcginnisIs there one?15:19
dhellmannthey're using falcon, pyramid, and paste?15:19
fungidhellmann: it looks like a cargo-cult list of deps to me15:20
*** kumarmn_ has joined #openstack-tc15:20
fungilike, they have eventlet in the requirements list but don't actually seem to import it anywhere15:20
* dhellmann wonders what "vine" is15:20
jrolldare I say twisted?15:20
jrollfor smcginnis, that is15:21
fungijroll: you, sir, are twisted indeed15:21
cdentsmcginnis: if you're running an api service, just don't; let the wsgi container run many threads. How many services are eventletting that actually need internal-to-the-process psuedo concurrency?15:21
mugsieor a ton of real threads and   let the kernal worry about it?15:21
cdentpy3 asyncio?15:21
jrollyes, this15:21
smcginnisHmm, I only have had a brief encounter with twisted, and can't really say I enjoyed it.15:21
dhellmannI would personally choose something that doesn't literally mess with the C execution stack at runtime15:21
ttxsmcginnis: i think we'd go full asyncio if we changed15:22
jrolldhellmann: ++15:22
ttxif only because its in stdlib15:22
smcginnisSo all the more reason to push for dropping py2 support.15:22
jrollthere's a library that attempts to backport asyncio to py2, AIUI, though I'm not sure how good of a job it does15:23
cdenthow much async of any kind do we actually need?15:23
*** kumarmn has quit IRC15:23
dhellmannjroll : I don't think haypo is supporting trollius any more15:23
jrollah, interesting, thanks15:24
dhellmanncdent : I could see it being useful in a few places where we're dealing with lots of data. It's less clearly useful for regular API calls, but I think another argument for adding it in the past was also the RPC layer.15:24
mugsiedhellmann: is it useful enough to offset the complexity it adds?15:25
cdentdhellmann: right, but doing things over RPC is another services inherit from nova without critically thinking about it15:25
dhellmannmugsie : I'm not advocating it, just trying to remember the arguments from the past15:25
cdents/another/another thing/15:25
dhellmanncdent : yes, true15:26
smcginnisMajority of usage I see in Cinder is just for sleep calls since IIRC time.sleep() has issues with the way we run. (?)15:26
jrollthe async model is useful for projects that spend lots of time waiting for hardware to do a thing (ironic, cinder, neutron, others?)15:26
smcginnisMost interesting usage in Cinder is with the backup drivers. Lot's of waiting on IO.15:27
dhellmannjroll : regular threads will block on I/O too, right?15:27
jrolldhellmann: sure, threads are one implementation of the async model15:29
dhellmannback to cdent's original question, did we answer it?15:29
cdentdhellmann: :) yes, thank you. And I actually think we moved on to a pretty good tangent.15:31
dhellmannyeah, I'd be interested to see what a new project would/could do with asyncio15:33
cdentI've added a bullet on
cdentas: why not?15:33
*** eumel8 has quit IRC15:33
openstackgerritPaul Belanger proposed openstack/governance master: Add naming poll info for S release
pabelangerttx: ^if we have some time to discuss today too15:35
dhellmannI think it's possible to run without eventlet. silhet and jd__ did some work in oslo.messaging a while back to support that.15:35
dhellmannso a service that expects it still needs it, but if the service isn't actually using eventlet it's possible to not use it and still use the lower layers.15:35
pabelangerit is possible dates might be too aggressive15:35
ttxpabelanger: no, sounds good to me. We can play the naming game all PTG week15:37
ttxIdem, if we can pile up approvals on this one I can approve it in time for you to start the process on the date indicated :)15:38
pabelangeryah, I'll start work on wiki pages today15:42
* dhellmann puts in an early vote for schnitzel15:43
ttxschadenfreude is too long15:44
fungiand not really a place15:45
smcginnisSome might see that as too close to home.15:45
fungid'oh! touché15:45
ttxwell schnitzel is not really German15:46
ttxspätzle, on the other hand...15:46
fungii was just about to suggest that ;)15:46
fungi(also very tasty!)15:46
ttxspandau is a Berlin district.15:47
fungispandau ballet was a briefly popular synth band in the '80s too15:47
mtreinishfungi: hah, I was just thinking of making a spandau ballet joke15:48
ttxSpeyer is one of the rare city names that fit under 10 characters15:48
smcginnisMe too.15:48
fungiyeah, deutsche is not known for brevity of its words15:51
persia"7254" is perfectly valid german, and considerably more concise15:51
fungibut doesn't 7 start with "z"?15:52
persiafungi: Unless I am out of practice, it is pronounced as smcginnis wrote.15:52
fungiyep, you're right. i mixed up pronunciation with spelling there15:52
pabelangerBased on previous nomination windows, if we started 2018-02-21, that will stay open until 2018-04-01 23:59. Are we good with that closing the thursday of PTG?15:52
pabelangererr, 2018-03-0115:53
ttxnah, I would wait the week after15:53
fungipersia: i was thinking of 2 which starts with "z" (10 does too i suppose)15:54
pabelangerttx: keep open day, and move close until 2018-03-07? that is 2 weeks to poll nominations, but we'd need to push back poll open to give time for foundation to review selection15:55
dhellmannI thought they only reviewed the selected items?15:57
dhellmannoh, you're talking about building the list, not voting, nevermind15:57
pabelangeryah, election officials need some time to finalize, then there is a week period before we open to voting15:58
fungiyeah, community proposes options, tc weeds out poor choices and presents a curated set, community votes on curated set ranking preferences, foundation  arketing and legal step through the preferred winners in sequence until they reach one which doesn't have to be stricken from the list15:58
pabelangerso, if we want to adjust the start time for nominations, they usually last a week. Which could be during PTG, or we can keep the nomination window open a little longer? I don't see any wording that is needs to be a week15:59
fungimainly because asking the foundation to weed out all contentious options is a wasteful cost of their time15:59
dhellmannpabelanger : do you have a pre-seeded set of names to start things off? or are we relying on people to add things to the list?16:00
pabelangerdhellmann: Good question, I need to see if mordred preseeding things him self, or depended on community 100%16:00
fungigood point. in the past having at least a few clear examples to start off the suggestion process helps tremendously16:00
fungito set the tone16:00
pabelangerbut agree, and example would be nice to have16:01
dhellmannmaybe one or two of the obvious cities or landmarks16:01
fungiyeah, otherwise the first entry in the list will likely be "snoopy"16:02
dhellmannhe did fly in WWI, right?16:03
fungiyeah, and his creator charles schultz has a german last name16:03
fungiso... win?16:03
dhellmann2 points at least16:03
persia might be a good place to start from for seeding16:04
fungiagreed. a sensible set16:04
fungii like salzburg16:05
persia and may also be useful16:05
fungii don't suppose we can have a release name with an umlaut16:06
pabelangerI'm going to push up an update for 545010 that will open voting 1 week later, to give some time for people to return from travels of PTG. That still gives us a window of 1 week after selection of names16:06
mordredpabelanger: I always just let the community do it16:06
fungithree cheers for openstack sülm!16:06
mordredpabelanger: but also, if you want to pre-seed, you are part of the community, so that's totally valid! :)16:06
openstackgerritPaul Belanger proposed openstack/governance master: Add naming poll info for S release
fungione option for pre-seeding is to not add explicit entries, but add links to things like the above lists16:07
fungias suggested places from which the community can draw ideas16:07
dhellmanngood idea16:07
pabelangermordred: ++16:07
fungido we explicitly require a 7-bit ascii representation?16:09
fungilike, strassberg instead of straßberg?16:09
pabelangertc-members: I've updated as to not close nominations during PTG. Please review again if you already voted.16:10
fungi(noting that ss isn't quite interchangeable with ß but works in a pinch)16:10
dhellmannfungi : yes, 26 characters of ISO basic Latin:
fungidhellmann: thanks for the reference!16:11
fungiso romanization possibly necessary for some selections16:11
fungihard to believe we're already talking about names for our 19th release cycle16:17
*** rosmaita has quit IRC16:28
*** hongbin has joined #openstack-tc16:30
*** kumarmn has joined #openstack-tc16:35
*** kumarmn_ has quit IRC16:38
*** ykarel has quit IRC16:46
*** harlowja has joined #openstack-tc17:19
*** cdent has quit IRC17:30
*** jpich has quit IRC17:32
openstackgerritJames E. Blair proposed openstack/governance master: Add resolution about CI for external projects
fungithis ^ would be quite helpful to the infra team17:49
*** kumarmn_ has joined #openstack-tc17:50
*** kumarmn has quit IRC17:53
pabelanger++ I'm in favor of supporting it. It does mean possible longer wait times for opentack projects to be tested. I know last PTG there were some people in the community that raised a voice when another project in openstack started consuming more nodes.17:53
pabelangerHowever, adding more cloud providers does help fix that :)17:54
dmsimardcorvus: left a comment17:56
*** rosmaita has joined #openstack-tc17:57
*** bmace has quit IRC18:02
*** dtantsur is now known as dtantsur|afk18:04
fungipabelanger: i doubt it means that in the near term. the quantity of third-party ci jobs we would run at the outset would be a tiny percentage of our overall job volume18:23
fungiif it becomes popular enough to represent a noticeable proportion of job volume (which i doubt it ever would) then there are plenty of options to address that when it happens18:24
funginot the least of which is that we can turn off our tests of external projects until we come up with a better solution18:25
fungiso i wouldn't want to see us get into the weeds predicting what that would look like18:25
pabelangerfungi: yup, agree. Just wanted to highlight a possible issue18:25
persiaAt this point, I think the only important part is to ensure the language permits that retraction (or other limitation) of service.  Until something happens, we can't know what will happen, but when it does, it is best to have a range of options.18:27
*** bmace has joined #openstack-tc18:29
*** persia has quit IRC18:39
*** harlowja has quit IRC19:00
EmilienMpabelanger: commented.19:04
EmilienMbut I'm not sure actually, let me verify again19:05
mugsieEmilienM: I am pretty sure Berlin was its own state19:09
EmilienMmugsie: I updated19:09
EmilienMbut yeah, Berlin is its own state I think19:10
mugsieah, that makes more sense for an area alright19:12
smcginnisI think we've preferred local areas, but allowed broader selection.19:13
pabelangerEmilienM: I don't think it is wrong, but reading just makes reference to a region encompassing the location of the summit. I took that to be Germany, but you are right, it is not a province / state but country19:15
EmilienMpabelanger: Ja, mein Freund!19:16
EmilienM(I didn't have to google it btw :P)19:16
EmilienMI suggest for any german question we poke cmurphy :) I've heard she's an expert now ;)19:17
cmurphyif the summit was in franconia i might have an idea19:17
cmurphyberlin is kind of a unique beast19:17
pabelangerWe could change it to the state of which Berlin is part of. I personally don't have a preference19:17
cmurphyberlin seems to be its own region and state19:19
pabelangerEmilienM: cmurphy: left a reply on 545010, if we want to update, I'm okay with that. But we should be more specific in our region selection (state / province)19:23
pabelangermordred: maybe you have a moment to look at 545010, since you've been responsible for naming releases until now^19:24
EmilienMhaving "Berlin" in region name works for me19:24
EmilienMso far this is what Wikipedia tells us19:24
mordredpabelanger, EmilienM: AH - so ... the process so far for this has been that I've just made a judgement call about what region 'feels right' and then proposed that19:25
cmurphyberlin is big enough and has enough history that there could be tons of ideas19:26
mordredyah. it's also conceivable that we'd have a second summit in germany in a non-berlin area19:26
EmilienMmordred: it's Ja, not "yah" (in german)19:27
* EmilienM runs19:27
mordredbut it's worth noting that catalonia, texas hill country and new england are not strictly speaking states/provinces19:27
mordredEmilienM: es tut mir leid :)19:28
openstackgerritPaul Belanger proposed openstack/governance master: Add naming poll info for S release
pabelangerBerlin it is!19:29
mordredpabelanger: lgtm!19:29
mordredalso - what with berlin being located on the Spree - MAYBE this time we'll actually pick a name that evokes the place in question19:30
* smcginnis sees a lot of Spree's candy at the summit.19:30
cmurphyyou don't think the havana release evoked portland?19:31
openstackgerritMatt Riedemann proposed openstack/governance master: Mark the mutable-config goal done for nova
* mordred is pretty sure we'll somehow figure out how to choose Strasbourg as the name19:31
mordredcmurphy: :)19:32
*** harlowja has joined #openstack-tc19:35
fungiabout as much as juno evoked atlanta19:36
EmilienMmordred: it's where I'm from19:36
fungior as much as liberty evoked vancouver for that matter19:37
*** harlowja_ has joined #openstack-tc19:37
openstackgerritSean McGinnis proposed openstack/governance master: Mark the mutable-config goal done for Cinder
*** harlowja has quit IRC19:39
*** hongbin has quit IRC20:31
*** hongbin has joined #openstack-tc20:32
openstackgerritJames E. Blair proposed openstack/governance master: Add resolution about CI for external projects
openstackgerritDoug Hellmann proposed openstack/governance master: replace tracking sections with storyboard links for rocky goals
*** openstack has joined #openstack-tc21:45
*** ChanServ sets mode: +o openstack21:45
*** cdent has joined #openstack-tc21:45
*** dtk has quit IRC21:58
*** dtk has joined #openstack-tc21:58
*** kumarmn has joined #openstack-tc22:50
*** kumarmn_ has quit IRC22:53
*** kumarmn has quit IRC23:25
*** cdent has quit IRC23:45
*** kumarmn has joined #openstack-tc23:51
*** kumarmn has quit IRC23:56
*** hongbin has quit IRC23:56

Generated by 2.15.3 by Marius Gedminas - find it at!