Monday, 2015-12-21

*** dims has joined #openstack-release00:31
dimslifeless : the change made its way into the images, so gate-nova-requirements in review ( passes now.00:35
dimslifeless : there are 3 *-constraints jobs that fail in that review. need to peel the next layer of onion00:36
dimslifeless : please take a look when you get a chance00:37
lifelessdims: are they passing on master?00:39
dimslifeless : we did not break any existing scenarios so we are good00:40
dimslifeless : in this review we are trying to use oslo.db[fixtures] instead of test-resources directly00:41
dims(for nova)00:41
dimsthis review used to fail in gate-nova-requirements before which we fixed00:41
lifelessdims: sure, but those jobs are non-voting, which usually means that they are failing on master // flaky00:41
lifelessdims: I wouldn't expect constraints to be flaky, so the question is whether they are passing on master00:41
dimslifeless : ah. they are passing on master00:41
lifelessdims: before we dig in deep00:42
dimslifeless : y, i have another review in parallel to test things -
dimsthe same 3 jobs work fine (green)00:42
lifeless2015-12-20 18:39:13.170 | ImportError: No module named testresources00:44
lifeless2015-12-20 18:39:13.170 | The test run didn't actually run any tests00:44
lifelessdims: I'm pretty sure pip isn't fixed yet00:44
lifelessdims: we're going to be looking at 3-6 months to fix this00:45
dimsall 3 jobs?00:45
lifelessdims: the pip patch is done, but we need it in a release, then time to get it out there00:45
lifelessdims: all three will have the same symptoms00:45
lifelessdims: pip sees oslo.db as a dependency and doesn't merge the extras in00:45
dimslifeless : one of them had a "Double Requirement" let me dig that up00:45
dimslifeless : ^^00:46
dimsso 2 problems i guess00:46
lifelessdims: same thing yes00:46
dimsok thanks lifeless. i can rest easy :)00:47
lifelessits closed, but not in a release00:47
dimsthanks lifeless00:53
dimslifeless : one more... do you want a pbr release this week?00:57
dimslifeless :
dimsplease open a review against releases/ repo if you do01:02
*** dims has quit IRC01:02
*** dims has joined #openstack-release02:07
openstackgerritDavanum Srinivas (dims) proposed openstack/releases: Oslo Releases for week of Dec 21st 2015
lifelessdims: no pbr release - not just before christmas03:17
dimslifeless :)03:19
*** dims has quit IRC03:32
*** amotoki has joined #openstack-release05:21
*** ifat_afek has joined #openstack-release06:19
*** amotoki has quit IRC06:59
*** ifat_afek has quit IRC09:02
*** ifat_afek has joined #openstack-release09:03
*** openstack has joined #openstack-release15:35
*** openstackstatus has joined #openstack-release15:36
*** ChanServ sets mode: +v openstackstatus15:36
*** stevemar_znc is now known as stevemar16:28
jrolldims_: we need a resolution on this before another o.msg release, or we're just going to break again
dims_jroll : good point16:33
dims_jroll : merging it16:34
jrolldims_: awesome, thank you16:34
openstackgerritMonty Taylor proposed openstack/releases: Release os-client-config 0.13.0
*** devananda has quit IRC17:08
*** devananda has joined #openstack-release17:15
openstackgerritDavanum Srinivas (dims) proposed openstack/releases: Oslo Releases for week of Dec 21st 2015
dims_jroll : ^^17:40
jrolldims_: thanks17:41
dims_mriedem : have a few mins to eye ball this?17:41
mriedemoh i suppose17:43
mriedemoh man, well, i'm going to have to check the git compare for the last release and see if there are any backward incompat changes being released17:44
dims_mriedem : look at the list-changes job output17:50
dims_mriedem : has all that good stuff17:50
mriedemooo that's nice17:58
mriedemi hope 789a6625980f97a302d6320d864cebb35b405e1c doesn't impact logstash18:02
mriedemi'm not sure that it would, i guess we'll know if it does18:02
*** dtantsur is now known as dtantsur|afk18:06
mriedemdims_: should have probably been emitting a deprecation warning on this option
mriedemi kind of wonder if removing that warrants a major version bump18:09
dims_mriedem : y borderline18:10
jrollmriedem: I agree completely18:12
openstackgerritSean McGinnis proposed openstack/releases: Release cinder 7.0.1 for stable/liberty
jrollidk what's borderline about removing a config option, especially when the default behavior was not the same as the behavior now18:12
*** Guest95009 is now known as jgriffith18:15
dims_jroll : please read the approved spec when you get a chance
jrolldims_: sure, but that still should be a major bump, and I wish it was deprecated properly18:17
dims_jroll : soon we'll be shipping 10000.1.018:18
jrolldims_: nothing wrong with that :)18:18
dims_if it helps folks from stopping using deprecated code, sure. it clearly does not18:19
dims_sorry, not in a good mood18:19
mriedemposted my comments, there were some suspect changes in taskflow and tooz also18:20
mriedemthat look like maybe a major version bump18:20
mriedemas for the oslo.messaging one, yeah removing deprecation options isn't in itself bad, but (1) the default behavior now is not the same and (2) it wasn't logging a deprecation warning when the option was being used18:21
dims_jroll : mriedem : at the moment the entire burden of this backwards stuff is on oslo team which is not fair18:21
mriedemwell, that's kind of how things work, right? things at the bottom of a software stack are slowest to be able to move18:21
*** electrichead is now known as redrobot18:21
mriedemb/c of deps up the stack18:21
dims_we get told not to break any stuff, but folks don't cleanup for several releases18:22
jrolldims_: if you give proper deprecation warnings for the proper amount of time, you should totally be able to break things18:22
jrollif people don't clean up their stuff, it's their fault18:22
dims_jroll : exhibit A - the timeutils + nova18:22
jrollI'm not up to speed on that one18:23
dims_jroll : my problem is that i do not see any effort towards a common goal18:23
dims_jroll :
mriedemdims_: those patches can be restored18:24
dims_mriedem : those are symptoms of the problem i am talking about matt.18:25
dims_be right back18:25
jrolldims_: I mean, the efforts are just beginning18:27
jrollbut really needing to do a major vs minor version bump isn't a burden at all18:27
dims_jroll : just complaining :)18:43
openstackgerritDavanum Srinivas (dims) proposed openstack/releases: Oslo Releases for week of Dec 21st 2015
dims_done. ^^18:45
dims_thanks for the reviews mriedem and jroll18:46
dims_mriedem : jroll : we totally missed any reference to configuration options in
dims_lifeless ^^18:47
lifelessdims_: ?18:50
dims_lifeless : was re-reading that doc to see if we cover "i am removing a deprecated option. Do i bump x or y in x.y.z"18:51
lifelessdims_: its an API break18:52
lifelessdims_: x, always x.18:52
lifelessdims_: (if config options are part of the public API)18:53
dims_lifeless : does it matter if the behavior remains the same after the option is removed or not?18:54
jrollafaik, it does not18:54
dims_jroll : not for this case, asking in general18:54
jrollI know :)18:55
dims_oh ok18:55
jrollbut I defer to lifeless anyhow, because he tends to know these things better than me18:56
lifelessdims_: might offer some clarity18:57
lifelessdims_: but in general think about it like a quantum physics problem... if the change is observable18:58
lifelessdims_: and its disruptive, then its not backwards compatible18:58
lifelessdims_: deprecating a thing is not disruptive (assuming folk have their log setups correct etc, which is a reasonable assumption)18:59
lifelessdims_: removing the deprecated thing itself is where the disruption occurs.19:00
lifelessdims_: also, another nuance - I believe for servers we don't consider config options public interface19:02
lifelessdims_: and instead only consider the HTTP API, though I suspect deployers don't really like that :)19:02
dims_lifeless : ack. trying to digest the nuances :)19:03
dims_mriedem : looks like harlowja responded on the review please take a look and i can cut the releases19:05
dims_after that19:05
mriedemdims_: i think taskflow probably needs a major version bump for the version field being dropped from the failure object, if that object is part of the public API for taskflow19:09
mriedemi think the version on tooz is probably OK as-is based on harlowja's reply19:09
dims_mriedem : let me drop taskflow from this review for now19:11
openstackgerritDavanum Srinivas (dims) proposed openstack/releases: Oslo Releases for week of Dec 21st 2015
dims_mriedem : ready for your +1 :)19:13
lifelessmriedem: so I think we have enough consensus on the backwards compat spec to start applying it - or at least raising it - on any backwards incompat change: "if it breaks supported servers, don't do it"19:16
lifelessmriedem: I haven't looked at the list of releases in detail, but any major version bump is a good alert :)19:17
mriedemlifeless: i noted this as needing a major version bump for oslo.messaging
mriedemi think that's the only one in dims_' current set of oslo releases19:19
dims_mriedem : right19:20
dims_lifeless : there19:26
dims_lifeless : there should be some incentive for folks to cleanup the deprecated stuff. no?19:27
*** doug-fish has quit IRC19:29
mriedemdims_: well, depends, that's like saying nova should have incentive to support glance v2 just because glance v1 is deprecated, but nova doesn't actually need to use anything in glance v2, or at least didn't use to (not sure if there is anything nova is depending on for glance v2 besides the fact v1 is deprecated)19:33
mriedemit was really the glance team that wanted that to happen, which is why they were/are driving it19:34
mriedemsame with keystoneauth19:34
dims_right, there is no incentive for the core team currently19:35
mriedemwell, one thing nova needs is an oslo liaison to raise these issues up
mriedemthat used to be jogo19:40
mriedemwe won't have a nova meeting until early january, but that should be on the meeting agenda19:41
dims_mriedem : i tried and haypo tried as well for a while after jogo left. at this point i believe for nova at least it should be a core who is a liaison. otherwise it wont work19:42
mriedema core liaison is probably best, sure, but honestly i don't know who has bandwidth for that right now19:43
mriedemi'll add it to the meeting agenda though19:43
mriedemherding cats and reporting status, attending oslo meetings, that doesn't need to be a core19:43
mriedemthat would actually be a good opportunity for a newer person looking to get involved19:44
dims_mriedem : that's exactly what we tell other teams19:44
dims_we can talk more about it in nova meeting. i am getting ready to push the release button19:45
mriedemlifeless: are you ok with the oslo.messaging major version bump here?
dims_mriedem : too early there i think :)19:48
mriedemnearly 9am19:49
dims_mriedem : i don't expect him to say no to the bump19:49
mriedemrelease if you want, it's getting late if you're going to do it today,19:50
mriedemsdague is out for 2 weeks i think,19:50
mriedemand i'm only working tomorrow19:50
mriedemso if everything explodes from this, i only have tomorrow to look really19:50
mriedemi'll assume that mtreinish is not taking any vacation time though since he's young and single19:50
jrollyoung single people deserve vacations too!19:52
mtreinishmriedem: I'm takng vacation, just not today19:57
mtreinishmaybe tomorrow, not sure yet19:57
openstackgerritMerged openstack/releases: Oslo Releases for week of Dec 21st 2015
mtreinishdims_: when you get a sec can you help with:
dims_mtreinish : upper-constraints already has tempest-lib===0.12.020:17
dims_mtreinish : want to rebase?20:17
mtreinishdims_: that was pushed before it got updated. We still need the g-r bump20:17
mtreinishsure, one sec20:17
* mtreinish notes the button works for that case20:18
dims_mtreinish : done20:18
mtreinishdims_: cool, thanks20:19
lifelessdims_: incentive yes, punishment no20:25
lifelessmriedem: dims_: looking20:25
lifelessmriedem: dims_ : Whats the incompatible change in messaging ?20:25
lifelessmriedem: so this breaks kilo [but the capped versions in kilo mitigate] ?20:27
lifelessactually, doesn't it break L too, per the docs ?20:28
lifelessmriedem: ^20:28
mriedemit would break anyone using that option in L i think20:28
lifelessso, yes, I object. We need to revert that, and leave it to L-eol.20:28
mriedemthe server projects don't use it as far as i know, but it can be passed through from the server config files20:29
lifelessremind me20:31
lifelessdoes oslo.config error on invalid settings20:31
lifelessbut foo isn't in the schema?20:31
lifelessor does it ignore them20:31
dims_lifeless : ignores20:31
lifelessthere are two parts here20:32
lifelessfirst, there's a bunch of cruft being cleaned up in the code20:32
lifelessand there is this option, which defaults *false*20:32
lifelesswhich we're now forcing to True20:33
dims_lifeless : this was the spec
lifelessso, is this wire level incompatible?20:34
lifelessits looking like it is20:34
lifelesswon't that break upgrades?20:34
dims_you can't mix juno and kilo20:35
dims_in the same install20:35
dims_if you want to use this new oslo.messaging library20:35
lifelessbut thats exactly how nova rolling upgrades work20:35
lifelessthey deliberately run during the rolling upgrade with half the stack old version, half the stack new version20:36
dims_lifeless : let's see if it shows up in a grenade failure20:37
lifelessso AIUI oslo.messaging @ time of L's release can handle either 1-part or 2-part messages20:37
lifelessdims_: lets work through it quickly here20:37
dims_lifeless : we'll probably need at least sileht for this20:37
dims_i don't want to guess20:37
lifelesswhen we change this, we'll always send 1-part, which means K cannot handle a message we emit20:37
lifelessL can handle it.20:38
lifelessBut, if this oslo.messaging is to be usable by L, and L is going to be able to upgrade from kilo, L still needs to be able to send messages to kilo mid rolling-upgrade, so L still needs this to be off by default.20:38
lifeless-> I think we can't change the default until K has eol'd.20:38
lifelessmriedem: ^ care to check my logic ?20:39
lifelessmriedem: since you spend way more time than me on stable issues...20:39
mriedemso the thinking is grenade in liberty will be broken by this20:42
lifelessif it ran unconstrained20:44
lifelessso specifically I'd expect requirements constraints updates to fail grenade after the release is done20:45
lifelessas I don't believe we have a job that reflects this yet20:45
lifeless(in L)20:45
*** bnemec has quit IRC21:26
*** bnemec has joined #openstack-release21:28
*** dims_ has quit IRC21:33
*** dims has joined #openstack-release21:33
*** dims has quit IRC21:43
*** nikhil_k has joined #openstack-release21:44
*** nikhil has quit IRC21:48
*** dims has joined #openstack-release21:49
openstackgerritMonty Taylor proposed openstack/releases: Release os-client-config 0.13.0
mordreddims: ^^ look, I made you a present :)22:07
* mordred hands dims a large box of kittens to go with it22:07
*** dims_ has joined #openstack-release22:08
*** dims has quit IRC22:08
*** dims has joined #openstack-release22:12
*** dims_ has quit IRC22:13
openstackgerritTony Breeds proposed openstack/releases: Release python-novaclient 2.23.3 for kilo
*** dims has quit IRC22:29
*** mriedem is now known as mriedem_away22:30
*** dims has joined #openstack-release23:21
*** dims has quit IRC23:30
*** dims has joined #openstack-release23:32
*** gordc has quit IRC23:42

Generated by 2.14.0 by Marius Gedminas - find it at!