Thursday, 2018-03-01

*** zaneb has quit IRC00:04
*** kumarmn has joined #openstack-tc00:19
*** kumarmn has quit IRC00:19
*** kumarmn has joined #openstack-tc00:19
*** kumarmn has quit IRC00:33
*** kumarmn has joined #openstack-tc00:34
*** kumarmn has quit IRC00:38
*** kumarmn has joined #openstack-tc00:46
*** kumarmn has quit IRC01:04
*** kumarmn has joined #openstack-tc01:04
*** kumarmn has quit IRC01:06
*** kumarmn has joined #openstack-tc01:06
*** Guest20946 is now known as mgagne01:10
*** mgagne has joined #openstack-tc01:10
*** ianychoi_ is now known as ianychoi01:26
*** kumarmn has quit IRC02:12
*** flwang1 has quit IRC02:21
*** flwang1 has joined #openstack-tc02:33
*** harlowja has quit IRC02:45
*** ykarel|away has joined #openstack-tc03:12
*** kumarmn has joined #openstack-tc03:55
*** kumarmn has quit IRC03:57
*** harlowja has joined #openstack-tc04:02
*** ykarel|away is now known as ykarel04:03
*** harlowja has quit IRC04:18
*** kumarmn has joined #openstack-tc05:58
*** kumarmn has quit IRC06:02
*** openstackstatus has quit IRC06:13
*** openstack has joined #openstack-tc06:15
*** ChanServ sets mode: +o openstack06:15
*** openstackstatus has joined #openstack-tc06:15
*** ChanServ sets mode: +v openstackstatus06:15
*** gcb has joined #openstack-tc06:46
*** lbragstad has joined #openstack-tc07:19
*** lbragstad has quit IRC07:28
*** gcb has quit IRC08:10
*** zaneb has joined #openstack-tc08:25
*** mriedem has joined #openstack-tc08:33
*** Guest87533 is now known as dansmith08:35
*** lbragstad has joined #openstack-tc08:36
*** bmace_ has quit IRC08:49
*** bmace_ has joined #openstack-tc09:05
*** zaneb has quit IRC09:06
*** zaneb has joined #openstack-tc09:11
*** zaneb has quit IRC09:16
*** gcb has joined #openstack-tc09:25
*** cdent has joined #openstack-tc09:30
*** rosmaita has joined #openstack-tc09:40
*** zaneb has joined #openstack-tc09:51
*** jpich has joined #openstack-tc09:51
*** ykarel has quit IRC10:06
*** ykarel has joined #openstack-tc10:06
*** lbragstad has quit IRC10:32
*** lbragstad has joined #openstack-tc10:41
*** gcb has quit IRC10:42
*** rosmaita has quit IRC10:47
*** gcb has joined #openstack-tc10:48
*** david-lyle has quit IRC10:57
*** ykarel is now known as ykarel|afk11:02
*** ianychoi_ has joined #openstack-tc11:06
openstackgerritMatt Riedemann proposed openstack/governance master: Add a resolution about stable branch EOL and "extended maintenance"
mriedemsmcginnis: tonyb: cdent: fungi: persia: ^11:07
mriedemttx: ^11:07
mriedemso let it be written, so let it be done11:07
cdentcool, noted, whereas11:07
*** ianychoi has quit IRC11:08
*** ianychoi_ is now known as ianychoi11:10
*** david-lyle has joined #openstack-tc11:10
*** rosmaita has joined #openstack-tc11:31
*** zaneb has quit IRC11:48
*** ykarel|afk is now known as ykarel11:52
*** jpich has quit IRC11:59
*** ykarel is now known as ykarel|away11:59
*** ykarel|away has quit IRC12:04
*** david-lyle has quit IRC12:06
*** bmace_ has quit IRC12:12
*** lbragstad has quit IRC12:13
*** gcb has quit IRC12:15
*** rosmaita has quit IRC12:16
*** bmace_ has joined #openstack-tc12:16
*** zaneb has joined #openstack-tc12:18
*** bmace_ has quit IRC12:18
*** david-lyle has joined #openstack-tc12:19
*** ykarel|away has joined #openstack-tc12:21
*** mriedem has quit IRC12:23
*** zaneb has quit IRC12:24
*** cdent has quit IRC12:30
*** gcb has joined #openstack-tc12:36
*** ykarel|away has quit IRC12:37
*** gcb has quit IRC12:37
*** gcb has joined #openstack-tc12:41
*** lbragstad has joined #openstack-tc12:42
*** gcb has quit IRC12:48
*** gcb has joined #openstack-tc12:50
*** rosmaita has joined #openstack-tc12:57
*** kumarmn has joined #openstack-tc12:59
*** gcb has quit IRC13:01
*** kumarmn has quit IRC13:04
*** cdent has joined #openstack-tc13:20
*** rosmaita has quit IRC13:25
*** mriedem has joined #openstack-tc13:28
*** gcb has joined #openstack-tc13:30
*** lbragstad has quit IRC13:31
*** bmace has joined #openstack-tc13:39
*** kumarmn has joined #openstack-tc13:43
*** rosmaita has joined #openstack-tc13:44
*** mriedem has quit IRC13:46
*** mriedem has joined #openstack-tc13:52
*** gcb has quit IRC14:02
ttxHey, tc-members! Whoever is at Croke Park Hotel, we'll meet Friday at 9:30am as planned14:23
fungiright on!14:24
ttxexact location tbd14:24
fungithat room is my favorite14:25
dtroyerI heard tbd was a bit cold this week…will bring a comforter from the room…14:26
TheJuliadtroyer: That is a great idea!14:43
smcginnisSo everyone knocks on ttx's room at 9:30?14:56
*** gcb has joined #openstack-tc15:03
*** diablo_rojo has joined #openstack-tc15:27
dtroyerthat's room 721 right?15:27
smcginnisQuestion for us to discuss at some point. The question came up in Cinder as to whether it is acceptable for an individual project to declare py2 support as deprecated, or if they would need to wait until we decide as a whole to deprecate it in OpenStack.15:29
cdentgood question15:29
cdentkind of a badge of honor15:29
*** gcb has quit IRC15:30
smcginnisMaybe. :)15:30
smcginnisBut really wondering if they would get scolded for not following the greater whole.15:30
cdentI would guess the critical factor is, as always, packagers.15:32
dtroyerIt seems to me to be the next step in visible progress for the eventual migration.  Indefinitely-long deprecation periods are usually not great, but at some point we need to take the next step.15:33
cdentBut since we need to declare lack of support pretty soon anyway, I would think the "deprecated" (as in there but expect it to leave) is what all projects should be saying15:33
dhellmannI was thinking the next step is to say "python 3 first" (assuming we're saying "python 2 first" now)15:34
dhellmannand to verify that we have all functional test and doc jobs moved over15:34
smcginnisIt's an interesting point right now where we have some distros that do not fully support py3 and some soon to be released ones that will not support py2.15:34
smcginnisdhellmann: I agree with that plan. But in addition to that, wondering if it is acceptable if a given project just says "python 3 only."15:35
dtroyerdhellmann: yes, I think those can run parallel.  I guess I'm saying projects that really are at the point where there is no more py3 migration work should go ahead and say so and default to py3 where possible15:36
dtroyerthat an artificial hold-back on deprecation may not be useful15:36
smcginnisI guess still not clear. Not just saying they support py3 and recommending py3, but doing things like removing six compat code and saying YOU MUST rather than YOU CAN or YOU SHOULD.15:38
dtroyerok, that is a bit harder15:39
cdentsmcginnis: probalby too soon I would guess15:40
dtroyerI understood it to be starting the clock to enable that in the near future15:40
cdentyeah, that15:40
dhellmanndtroyer : right, I'm differentiating between "by default" and "only"15:40
smcginnisIt would be starting that clock, but jumping out ahead of the rest.15:41
dtroyerunless we are planning/hoping for a py2 drop-dead reprieve, we have a hard deadline approaching to meet our deprecation timelines15:42
dtroyerI am ok starting those clocks as projects are ready and dealing with the exceptions.  aside from distros dropping py2, do we have issues running mixed-mode here (cinder on py3, swift on p2 for example)?  Also assuming no containers that would allow that easily15:44
smcginnisI do see more and more deployments opting for containerized deployments, so that really does make it a non-issue.15:45
smcginnisIt's only an issue really if they want to install cinder and swift on the same host, but that's not a very good idea in the first place.15:45
dmsimardsmcginnis: depends -- kolla has centos based containers for example which would not work (right now) if only py3 was available15:46
dtroyerright, just an example.  but what prevents that besides the deployemnt tooling?15:46
smcginnisdtroyer: Can you clarify "that" in "what prevents that?"15:48
dtroyersure.  bettter example:  is there anthing in the projects themselves that would prevent nova-compute running on py2 and cinder-volume running py3 on the same node at the same time?  The issues are in deployment tooling/packaging, not in project deliverables?15:49
smcginnisI think maybe the right answer is first all projects MAY BE run py3 only, then once we've passed that bar projects are open to decide to change that to MUST BE.15:49
smcginnisdtroyer: Yes, I believe that is a correct statement.15:49
dtroyerI recall attempting to make this possible in DevStack and that we stopped short of that for $REASONS15:49
dtroyernot having to do with the projects themselves15:50
smcginnisOne of the $REASONS I was given was swift not being able to be run py3, so that would mean dropping devstack support for swift.15:50
dtroyerand basically that's where I'm heading here.  We hold back everyone until Swift and whoever else gets around to finishing py3 work?15:51
smcginnisI think we may need to. Unless we want to complicate things like running devstack deployments for testing.15:51
smcginnisWell, unless devstack can figure out a way to handle that.15:52
dtroyerI don't see running DevStack as a blocker to Cinder being able to say py2 is deprecated, starting the clock to when support is acutally removed15:52
smcginnisOh, true. It could be declared deprecated, just not removed.15:52
dtroyerDevStack can figure that out, its just messier than it already is.  The answer was going to be to put each service into a venv15:53
smcginnisMight be a good recommendation for all projects that are ready to start declaring it deprecated.15:53
smcginnisThen when swift and whatever else is ready, those projects that are ready and have deprecated it can start actually removing compat code.15:53
dtroyerright, that's what I am thinking too15:53
smcginnisThat seems like a reasonable approach.15:54
notmynamewhat sort of timeframe is in your mind?15:54
smcginnisnotmyname: You have any plans for next week?15:55
EmilienMttx: ack, I'll try to join 9am15:55
notmynamethat's ... a bit more aggressive than I was thinking ;-)15:55
smcginnisnotmyname: Is the work scoped enough that declaring something like T would be reasonable?15:55
dhellmannfrom what I've seen swift is still blocked on some upstream issues15:56
dhellmannunfortunately I didn't have time to catch up with notmyname before the evacuation15:56
smcginnisAnd to be fair, if Cinder decides to call it deprecated now, we would not necessarily state when it would be expected to be removed. Just that, hey, this is going away in the nearish future so be ready for that.15:56
notmynamemany of the problems are identified. they're not scoped, per se, but I don't think there are too many dependency issues15:57
dhellmannsmcginnis : are every single one of cinder's jobs running py3 now, aside from the py2-specific ones? docs? functional tests?15:57
smcginnisI don't believe docs are at the moment. But that's kind of secondary in my mind.15:58
notmynamesmcginnis: the T release is ... we're starting R now? so that's 18 months?15:58
smcginnisUnit tests and functional tests are running py3.15:58
dhellmannnotmyname : the stuff associated with internals of the socket module caught my eye as potentially challenging to plan a "schedule" around fixing. Are those being actively worked on? Do we need help from upstream python to add APIs to replace the use of those internal bits?15:58
dhellmannsmcginnis : at some point soon it's going to be hard for us to run py2 on a modern node in the gate, so it's pretty relevant15:58
smcginnisThe biggest cocern at this point is third party CI is not running py3, so we are not sure if there are any hidden driver issues.15:58
dhellmannand we have doc tool dependencies we may not have looked at ensuring work with py315:59
smcginnisnotmyname: Yeah, 18 months I think.15:59
notmynamedhellmann: I don't know. but TBH I hadn't ever really considered that an option15:59
dhellmannsmcginnis : that's another good point15:59
dtroyercan we work backwards from when we (OpenStack) feels the drop-dead on py2 is?  or do we plan to keep best-effort as long as the distros we use allow?15:59
dhellmannnotmyname : it's worth discussing if there's a valid use case. it might be rejected, but there's no reason *assume* it will be16:00
smcginnisdhellmann: Yeah, docs tooling needs to be there, but I think we can run those under py2 and still expect users to deploy the services under py3.16:00
dhellmann"we were doing this and can't any more, how do we do that now?"16:00
dhellmannmaybe there's an alternative, maybe we make one16:00
dhellmannsmcginnis : distros that don't support py2 still want to package docs16:00
smcginnisdtroyer: Good idea. Looking at 2020 as a deadline, T might be a good target.16:00
notmyname"we were using internal variables before and you moved them!" isn't normally something that goes over well :-)16:00
dhellmannI mean, docs are going to be the least troublesome part of the port. let's just not ignore it.16:00
dhellmannnotmyname : no, but if it's framed as "we need an API to do X and there isn't one" then it would be the start of a discussion16:01
smcginnisdhellmann: Yeah, definitely needs to be addressed.16:01
notmynamedhellmann: yeah, IIRC we ended up using them because we needed access to the actual socket object. it was to fix socket leaks in some conditions. so there may be valid justification16:02
*** gcb has joined #openstack-tc16:02
dhellmannnotmyname : yep, that sounds like something we could build a good case around. I'm happy to help you find the people you need to talk to, as well.16:03
dhellmannI think one of the cases was adjusting some sort of buffer size or other behavior like that, and that feels like a missing API. If the other was a bug, that seems like an even easier thing to sell them on.16:04
smcginnisWhat if we rewrote parts in go?16:04
* smcginnis ducks16:04
* dhellmann lobs a snowball at smcginnis 16:04
dtroyersmcginnis: what if that already happened?16:05
notmynamesmcginnis: honestly, since we are/were planning on doing that, that is a valid thing :-)16:05
dhellmannsure; that's an alternative that lets us say we're not blocked on a python 3 port16:06
smcginnisThis might be a more compelling reason for that to happen.16:07
notmynamesmcginnis: just to be clear, are you suggesting that the T release has py3 support for everything (and py2 is depricated)? or that py2 is no longer used in the T release? and do you mean at the end of the T release or by the time the T release starts?16:10
notmynameI assume "py2 is depricated" means that py2 isn't run in the gate obs at all16:11
dhellmannhmm, I'm not sure I assume that, so we should definitely agree on exactly what we mean16:13
smcginnisnotmyname: We may still want tests, but I was thinking T could be the point where all services are at least capable of being run with py3, and if a given project so chooses they would then be able to remove the six and all that compatibility code and at least declare for that project they only support py 3.16:13
dhellmannI'm much more concerned with the gaps identified in
dhellmannT might be a good release to do that; there's still a chance it will be 8 months instead of 6 if we don't have a PTG a year from now16:14
dhellmann(pending a decision from the foundation and bod)16:14
*** gcb has quit IRC16:15
notmynamesmcginnis: and that would be at the start or by the end of T?16:15
smcginnisBy the end of T I think.16:15
notmynameah, right. that's what dhellmann just said :-)16:16
notmyname"all openstack projects can at least run in both py2 and py3 by the end of the T release (expected roughly 18 months from now)" seems to be possible16:25
*** gcb has joined #openstack-tc16:31
*** gcb has quit IRC16:37
dhellmannwe're pretty close to that already, so yeah16:41
*** cdent has quit IRC16:46
*** rosmaita has quit IRC16:56
*** lbragstad has joined #openstack-tc16:57
*** mriedem has quit IRC17:02
*** diablo_rojo has quit IRC17:05
*** rosmaita has joined #openstack-tc17:06
fungilet's rediscuss tomorrow, but sounds good to me17:09
*** lbragstad has quit IRC17:10
*** lbragstad has joined #openstack-tc18:03
*** lbragstad has quit IRC18:06
*** lbragstad has joined #openstack-tc18:06
*** MatrixBridge has joined #openstack-tc18:11
*** MatrixBridge has left #openstack-tc18:11
*** gcb has joined #openstack-tc18:17
*** rosmaita has quit IRC18:24
*** lbragstad has quit IRC18:55
*** harlowja has joined #openstack-tc20:27
dhellmannare we still planning to try to meet friday?21:21
smcginnisdhellmann: 9:30 tomorrow morning. Location TBD.21:46
*** rosmaita has joined #openstack-tc22:25
*** kumarmn has quit IRC22:44
*** kumarmn has joined #openstack-tc23:03
*** kumarmn has quit IRC23:05
*** kumarmn has joined #openstack-tc23:05

Generated by 2.15.3 by Marius Gedminas - find it at!