Wednesday, 2018-02-07

openstackgerritJeremy Stanley proposed openstack/governance master: Remove editor mode setting from goals template
openstackgerritJeremy Stanley proposed openstack/governance master: Rename goals/template.txt to .rst
dhellmannfungi : I think you're going to want to add the file to the exclusion list to make sphinx not scan it and then report an error because it's not in toctree list ^^00:04
fungidhellmann: yeah, likely. i figure if i push it up i'll find out quickly00:04
* dhellmann nods00:04
fungiwell, for backlogged ci versions of quickly at any rate. it's not an urgent change, just annoyed at seeing editor configuration copied around likely by people who don't even use the editors in question00:05
fungipossibly faster than i'll figure out what libpcre package the docs build needs me to install on my workstation at any rate00:09
fungino idea why python-pcre can't provide a manylinux1 wheel00:11
fungimaybe because its last release basically predates manylinux100:11
* fungi sighs00:11
fungiof course, i was missing libpcre3-dev. i wonder if we should add a bindep.txt to openstack/governance00:14
fungiValueError: No project 'Packaging-deb' in projects.yaml00:15
fungiso why is it even finding Packaging-deb?00:17
fungithe change to move it into legacy merged back in late december00:18
*** kumarmn has joined #openstack-tc00:23
fungimmm, looks like i may have been in need of a thorough git clean there for a while00:25
fungifinally able to confirm00:25
fungisphinx.errors.SphinxWarning: /home/fungi/work/openstack/openstack/governance/doc/source/goals/template.rst:: WARNING: document isn't included in any toctree00:25
fungii'll add it to the hidden one00:25
*** kumarmn has quit IRC00:29
openstackgerritJeremy Stanley proposed openstack/governance master: Rename goals/template.txt to .rst
openstackgerritJeremy Stanley proposed openstack/governance master: Remove editor mode setting from goals template
*** hongbin has quit IRC00:35
fungii find it mildly interesting that and are so similar, proposed by different people from the same employer a couple weeks apart00:49
fungiall the proposed changes in gerrit from those two contributors fall into the same category00:51
fungionly a few dozen in total though, and ostensibly valid corrections/cleanup00:52
fungibut all within the last few weeks00:52
fungiskimming the commits stackalytics references for the top contributors from the same member company, they all seem to be that way00:57
fungidiablo_rojo: ^ probably one for the first contact sig, if they've not already been engaged?00:58
smcginnisfungi: I've seen a few of those patches go through.01:00
smcginnisNow that the Great HTTP Purge of 2017 is done.01:00
fungiand now i guess it's office hour time01:01
smcginnisWow, I made it to one of these!01:01
fungipurely by accident i'm sure. or at least you can try to tell yourself that01:02
smcginnisNot a lot to discuss with the two of us.01:07
fungimaybe some interested apac contributors will show up01:10
fungithis is supposed to be the apac-friendly timeslot, after all01:11
mtreinishfungi: it's not just apac friendly. I'm normally online this time of day too :p01:24
fungisure, americas night owl friendly too then01:24
* fungi notes he too is usually around01:24
fungistill trying to catch up on all the things i shrugged off earlier today to deal with infrasplosions instead01:25
*** kumarmn has joined #openstack-tc01:27
*** liujiong has joined #openstack-tc01:33
*** liujiong has quit IRC01:45
*** liujiong has joined #openstack-tc01:46
smcginnisfungi: Have the embers died down?01:50
fungifor the most part01:50
smcginnisCheck queue is still high, but at least it looks like it's not growing anymore.01:51
fungi says our node requests are finally down to ~1k (which is approximately our aggregate quota) so the backlog should hopefully clear in the next few hours01:52
fungibasically we need to turn over all our nodes once to fulfill the current backlog and this time of day we don't tend to add much01:52
fungiso estimating average job runtime at an hour, that's about how long it'll be before all runnable jobs are running01:53
*** kumarmn has quit IRC02:04
fungianyway, i think that's it for me tonight. i'm going to slowly poke at some election tasks but basically down for the count02:13
smcginnisMe too, see you fungi.02:14
*** mriedem has quit IRC02:24
*** kumarmn has joined #openstack-tc02:30
*** kumarmn has quit IRC02:45
*** kumarmn has joined #openstack-tc02:47
*** kumarmn has quit IRC03:07
*** kumarmn has joined #openstack-tc03:12
*** kumarmn has quit IRC03:24
*** coolsvap has joined #openstack-tc03:36
*** kumarmn has joined #openstack-tc03:37
*** ianychoi_ has quit IRC03:49
*** ianychoi_ has joined #openstack-tc03:50
*** kumarmn has quit IRC03:57
*** kumarmn has joined #openstack-tc03:57
*** openstackgerrit has quit IRC04:04
*** kumarmn has quit IRC04:04
*** kumarmn has joined #openstack-tc04:05
*** kumarmn has quit IRC04:09
*** rosmaita has quit IRC04:14
*** ykarel|away has joined #openstack-tc04:30
*** harlowja has quit IRC04:32
*** eumel8 has joined #openstack-tc04:59
*** openstackgerrit has joined #openstack-tc05:00
openstackgerritMasayuki Igawa proposed openstack/governance master: Add Cold upgrades capabilities
*** harlowja has joined #openstack-tc05:37
masayukig^^^^^^^^ EmilienM and TC guys, I updated the goal. Could you take a look when you are available?06:11
*** harlowja has quit IRC06:52
*** ykarel|away is now known as ykarel07:07
*** pabelanger has quit IRC07:50
*** pabelanger has joined #openstack-tc08:02
*** ianychoi_ has quit IRC08:14
*** ianychoi_ has joined #openstack-tc08:19
*** ianychoi_ has quit IRC08:28
*** ianychoi_ has joined #openstack-tc08:30
*** liujiong has quit IRC08:39
*** ianychoi_ has quit IRC08:45
*** ianychoi_ has joined #openstack-tc08:46
*** jpich has joined #openstack-tc08:59
*** kumarmn has joined #openstack-tc09:05
*** cdent has joined #openstack-tc09:06
*** kumarmn has quit IRC09:10
*** cdent has quit IRC09:36
*** cdent has joined #openstack-tc09:38
*** ykarel is now known as ykarel|lunch09:54
*** ykarel|lunch is now known as ykarel10:20
*** ianychoi_ has quit IRC10:21
*** ianychoi_ has joined #openstack-tc10:24
*** dtantsur|afk is now known as dtantsur10:26
*** gcb has quit IRC10:29
*** cdent has quit IRC10:36
*** ianychoi_ has quit IRC11:05
*** ianychoi_ has joined #openstack-tc11:07
*** cdent has joined #openstack-tc11:20
cdentttx: I'm +1 on time for the topic "where does the open source upstream project stop"11:55
cdentand tuesday afternoon is good11:55
*** ykarel has quit IRC11:59
*** ykarel has joined #openstack-tc12:00
*** ianychoi_ has quit IRC12:00
EmilienMmasayukig: sure12:52
ttxLooks like the mox goal is facing some resistance on the review... We could substitute the cold upgrade capabilities one to it.12:56
cdentttx there was some interesting conversation in the nova channel about that yesterday evening. Most of it is reflected in the comments that were left on the review, but a thing that stood out a bit more strongly in the conversation was "why don't we just keep mox3 alive?"12:59
ttxThat's a fair question and it made me pause... I think it's a question of code surface and if you consider that tech debt, the need to reduce it.13:03
ttxBut clearly if the cost of "maintaining" mox is limited, that drops the priority of gettign rid of it13:04
ttxSo I think that's the missing bit of information. How costly is it to keep it going.13:05
*** rosmaita has joined #openstack-tc13:06
cdentPerhaps, but I also think that "we're doing a non-standard thing" is reasonable reason to change.13:10
persiaUnless that non-standard thing is intentional, and a result of considered opinion.  For example, I think gerrit+zuul is incredibly good, but it's something we do for opinionated reasons, and is far from standard.13:18
cdentpersia: "non-standard" was probably the wrong phrase. Perhaps "non-idiomatic". In the sense that 'mock' is the way people mock in Python, generally.13:19
persiaI have a much easier time agreeing with that phrasing :)13:19
cdentI'm certainly not of the opinion that everybody should use the same thing if there are lots of different good things. zuul's great. I also happen to like gabbi. Neither are in a domain that is sufficiently rife with idioms for that to be an issue.13:20
cdentgerrit on the other hand... I really think gerrit is an actively damaging tool, but like so many tools is the least bad.13:21
persiaBecause of the terrible UI, or because of the workflows it imposes?13:21
cmurphycdent: wow that's an interesting take13:21
*** kumarmn has joined #openstack-tc13:22
cdentoh noes, have I made a "hot take"? quick, to twitter! ;)13:22
cdentpersia: it's hard to explain but short answer "workflows". Slightly less short answer: the impact it has on the human interactions that surround the review process. Very easy to tunnel vision or lose track of the larger points of code review in favor of details.13:24
cdentof course the ui _is_ terrible, and helps to drive the workflow issues13:25
persiaAh, makes sense.  If there is a pause, I'd be delighted to discuss more in Dublin, as I find many of the other available workflows even more damaging, and would like to have a clear vision of the "right" tooling.13:25
dimscdent : do we have outstanding requests on stuff to be fixed in mox3?13:25
cdentdims: as far as I'm aware (which is not very) I don't think there is much broken. It's stable. Does what it says on the tin. which is why some people are saying "let's just carry on then"13:27
*** kumarmn has quit IRC13:27
dimsjust not add any more was the line we used to take before13:27
cdentpersia: yeah, other workflows often even more damaging, thus  the "least bad". Also, as you're aware, there's a whole complex of factors involved in the experience of creating code and it's unclear how much of gerrit is a cause and how much is it simply helping things there anyway13:29
cdentdims: as I said in the nova conversation last night, I'm not wed to anything with regard to mox. My engagement in the topic is because I want the goals process to be real and useful and have a style of authority13:30
persiaYep.  That's why I'm interested.  In other contexts, I end up recommending Gitlab or Gerrit or Gitano or whatever, and I'm wondering if it is possible to make a tool that avoids the traps.13:30
fungicdent: i tend to agree with you somewhat on gerrit. i still slightly prefer th lkml workflow and mailing lists for code review, but have to admit gerrit brings quite a few efficiencies (granted, at the cost of a slight decrease in holistic approach to review)13:30
fungibut... at least it's not github!13:30
cdentgithub is nice for small things, but falls apart otherwise13:31
cdentI sometimes react to that thinking: everything should be small13:31
fungii accept gerrit as a necessary evil because i doubt we'd achieve similar throughput via e-mail13:31
cdentfungi: yes13:32
cdentIt's certainly the case that a lot of not-great stuff which could be attributed to gerrit could just as easily be attributed to the social pressures of scale13:32
fungibut gerrit does definitely focus the reviewer more on line-by-line feedback than discourse13:33
fungiit has facilities to support both, yet seems to attract/encourage more of the former than the latter (particularly due to lack of discussion threading)13:34
cdentyes and yes13:34
cdent"discourse" is probably a key word in this topic13:35
cmurphycomments have a little reply button now that does quoting13:35
fungion lkml you frequently get good discussions of an entire patch series rather than just patch-by-patch (and a lot less quibbling over a single line)13:35
fungicmurphy: sure. the quoting dos at least help provide context in busy discussions. it's no substitute for threading but better than nothing13:36
fungiand we do still get those broader discussions as well, but end up falling back to our mailing list to have them13:37
persiaDuring some of the pynotedb discussions, the gerrit folk were open to the idea of dealing with series, rather than individual patches, but it would require a UI overhaul.13:41
*** coolsvap has quit IRC13:41
cdentcmurphy: would you describe yourself as a "fan of gerrit" or "accepting of gerrit" or something else?13:50
cmurphycdent: fan of gerrit13:53
cmurphyhaving come from github13:53
cdentbig projects on github?13:53
cmurphyit encourages small self-contained changes13:53
cmurphyi'd say medium-size and small13:54
cmurphypuppet modules and puppet13:54
cdentI agree that the way you can separately evaluate a series of related changes in gerrit can be very nice, but it can also be very hard to see (or talk about the big picture)13:55
cdent(btw: I'm not suggesting we change things or anything like that, just making conversation)13:55
cmurphyi guess i'm a fan of the fact that we have multiple communication channels and we don't rely on only our code review tool to talk about things13:56
fungitrue, i think we do a good job of balancing discussion between multiple venues13:57
cmurphycode review tools are for code review, mailing lists are for async communication, irc is for sync communication13:57
cmurphytrying to do every part of open source through github is trying to assign too much responsibility to one tool13:57
fungithe line between "code review" and "discussion" can be blurry though13:58
cmurphyand discussions overlap13:58
cdentI can sense a moment of violent agreement approaching13:58
* cdent braces self13:58
fungicmurphy has that effect on people13:58
fungi"stop making me agree with your salient points!"13:59
fungi( what i would be saying, but now i need to do some code review)14:00
* TheJulia sips coffee and reads14:14
*** ianychoi has joined #openstack-tc14:17
*** kumarmn has joined #openstack-tc14:22
*** david-lyle has quit IRC15:15
*** coolsvap has joined #openstack-tc15:17
*** mriedem has joined #openstack-tc15:26
*** hongbin has joined #openstack-tc16:07
*** ykarel has quit IRC16:14
*** david-lyle has joined #openstack-tc16:18
*** eumel8 has quit IRC16:46
*** kfox1111 has joined #openstack-tc17:11
kfox1111question... does all projects support osprofiler yet? If not, would that be a good cross project goal?17:11
*** harlowja has joined #openstack-tc17:11
kfox1111our openstack dashboard has been getting slower and slower and we're starting to have to instrement things to figure out how to fix it.17:12
*** mriedem has quit IRC17:14
*** mriedem has joined #openstack-tc17:15
dhellmannmox3 is currently listed as maintained by the oslo team, but we wanted to stop having to maintain it. If we're going to keep it, we probably want to find other owners for it, or at least ask the oslo team if they will continue to manage it. I've put that on the agenda for the Oslo team at the PTG.17:15
*** cdent has quit IRC17:17
*** harlowja has quit IRC17:17
rosmaitasorry to interrupt ... is it still office hours?17:18
dhellmannrosmaita : not really but that's no reason not to bring something up. what's on your mind?17:19
dhellmannkfox1111 : I don't think all projects do support it yet, but I'm not sure. That might make a good goal. Are you interested in doing the work to figure out what would need to be done?17:20
rosmaitawhat's our current position with respect to postgres?  i.e., suppose there was a hypothetical openstack project that didn't want to support it any more, could that project just stop?17:20
dhellmannthat came up at the summit in boston I think17:21
openstackgerritSean McGinnis proposed openstack/governance master: Retire horizon-cisco-ui project
dhellmannwe have a couple of big distributors and/or users who are relying on it, so I think we said we wouldn't drop it based on that meeting17:21
dhellmannlet me see if we wrote that down somewhere17:21
dhellmannrosmaita :
*** cdent has joined #openstack-tc17:22
cmurphyfor the record suse is working toward becoming not one of those distros17:22
kfox1111dhellmann: not sure. depends on if I need to do the work to fix the problem.17:22
dhellmanncmurphy : ++17:22
rosmaitadhellmann ty17:22
kfox1111just starting to look into osprofiler.17:22
dhellmannI think the other big company using it was Huawei? but they had their own fork of PG or something?17:22
kfox1111would really like to hook it into something like jaeger so it has less overhead.17:22
dhellmannsmcginnis, dims : do you know about postgresql support at Huawei? ^^17:23
mriedemhuawei uses guassdb17:23
mriedemwhich is a pg fork i think17:23
smcginnisdhellmann: I believe our product still uses a postgres.. yeah, that17:23
dhellmannkfox1111 : that sounds like an interesting project. what would Jaeger replace? (I don't know what osprofiler is using as a backend these days)17:23
kfox1111dhellmann: currently osprofiler sends everything through ceilometer.17:24
mriedemSAP also uses pg i think17:24
dhellmannsmcginnis , mriedem : it sounds like the glance team is interested in being able to drop support. there may be an opportunity to contribute there17:24
kfox1111which is kind of a pain/slow. and profiling everything is costly.17:24
kfox1111last I looked at jaeger, it had the option to at runtime capture only some traces while ignoring the rest.17:24
kfox1111so much better to leave in place in production.17:25
mriedemwhat is glance trying to do that won't work with pg over sqla?17:25
dhellmannrosmaita : ^^17:25
kfox1111I saw a ps adding jaeger support to osprofiler.17:25
rosmaitasmcginnis mriedem dhellmann : our rolling upgrade stuff17:25
kfox1111hasn't merged yet. not sure if its usable or not.17:25
mriedemcan you make it optional?17:25
mriedemi.e. if you want rolling upgrades, use mysql, else you don't get rolling upgrade17:25
mriedemare you talking about db triggers or something?17:25
smcginnisWhat is the issue with rolling upgrade and postgres?17:26
mriedemso if engine == 'mysql': do that thing17:26
smcginnisrosmaita: Just looking at some mysql optimization?17:26
mriedemi assume it's going to use native sql rather than sqla ORM?17:26
rosmaitayeah, but right now it's if engine != mysql barf17:26
mriedemso make barf == offline data migration17:27
dhellmannare we already using that pattern elsewhere?17:27
rosmaitanot sure17:28
mriedemnova does online data migration in code17:28
mriedemon read17:28
mriedemdata migration is never done during nova-manage db sync17:28
mriedemand we have CLIs to let you run those in batches as you like17:29
mriedembeen doing that since kilo17:30
dhellmannyeah, I think that's the preferred way to do it, but I was asking if we already have another project doing the online/offline mode switch based on the database17:31
dhellmannmaybe keystone?17:31
mriedemno lbragstad here...17:31
dhellmannI remember lots of discussion about not using database triggers or stored procedures and keystone going ahead anyway. rosmaita, you might consider the python-code approach mriedem is describing as an alternative.17:31
rosmaitai think ours was roughly based on keystone and maybe neutron17:32
*** dtantsur is now known as dtantsur|afk17:33
cmurphykeystone is using triggers17:33
cmurphyi don't know the history there though17:33
*** jpich has quit IRC17:37
*** harlowja has joined #openstack-tc18:11
*** mriedem has quit IRC18:14
*** harlowja has quit IRC18:18
openstackgerritMonty Taylor proposed openstack/governance master: Dedicate Queens release to Shawn Pearce
mordredttx: ^^ fixed19:35
*** mriedem has joined #openstack-tc19:35
mordredcmurphy, rosmaita: I was one of the voices of "please dear god do not use triggers"19:36
mordredcmurphy, rosmaita: I got ignored because "we already wrote the code" IIRC19:36
mordredtriggers are a TERRIBLE idea if you want things to be scalable - as they locate CPU logic and activity in the portion of the infrastructure that is the hardest to scale19:37
mordredthey're also a TERRIBLE idea from a maint perspective as long as we support more than one database engine19:37
mordredgiven that all of the SQL written in openstack is doing through the SQLA abstraction, having some actions all of a sudden be in direct low-level SQL is conginitively dissonant19:38
mordrednow - ignoring the scaling and debugging concerns, if we wanted to shift our SQL layer to being SQL statements and not ORM generated it MIGHT make more sense, because then we'd develop tooling around validing SQL19:39
mordredbut - triggers are terrible from a scaling perspective and are also a completely different beast for an admin to debug19:39
mordredsince errors in a trigger are going to go into database logs and will not show up in openstack operational logs19:39
mordredanywa - that's my soapbox about triggers and stored-procedures in modern web apps19:40
mordredthey were a fantastic idea when the world tended to be one very large and expensive machine with the Oracle database on it and a wealth of applications that used the storede procedures as an API access layer to the db19:40
persiaSPoT scaling is CAP-limited, so yeah, even if 4GLs are "better", they don't scale.19:40
persiaIt also works with other-brand DBs :)19:40
mordredbut that hasn't described the state of the world for the web in MANY years19:41
mordredpersia: ++19:41
*** coolsvap has quit IRC19:41
mordredthis is, incidentally, the same issue I have with actually applying foreign keys at the DB layer - it's a solution to a problem written for a time we no longer live in19:41
mordredand although it's 'correct' from a relational algebra perspective, it's *wrong* from a "I want this bad boy to actually work at scale" - when all the access to the db is going to be flowing through an app layer anyway19:42
mordredBUT - I've given up on selling people about that19:42
persiaI think of constraints as useful for developers, in terms of causing early errors in CI.  The only reason to have them in production is the convenience of identical code for production and testing.19:43
*** jpich has joined #openstack-tc20:04
openstackgerritJeremy Stanley proposed openstack/governance master: Rename goals/template.txt to .rst
openstackgerritJeremy Stanley proposed openstack/governance master: Remove editor mode setting from goals template
*** kumarmn_ has joined #openstack-tc20:37
*** kumarmn has quit IRC20:41
*** jpich has quit IRC21:04
*** mriedem has quit IRC21:53
dimsfungi : just sold yaml2ical to sig-contribex folks ( ) hurray21:54
fungitonyb: ^ your baby's all grown up now21:55
fungioh, i guess lbragstad and ttx did the bulk of stuff in there after the initial intern efforts trailed off21:58
fungiand mrodden21:59
tonybdims: \o/22:02
tonybfungi: Yeah I did only small amounts22:02
dimsdon't know if they will pick it up. at least made sure it was on their radar :)22:02
tonybfungi, dims: I knwo it was used in centos and probaly still is22:02
tonybdims: well if they do or they don't for silly things like missing features I'm happy to work with $people to address that.22:03
dimsack tonyb ! will report back if there's any progress22:03
dimsand good morning :)22:04
tonybdims: afternoon/evening ;P22:06
*** cdent has quit IRC22:22
*** kumarmn_ has quit IRC23:38
*** kumarmn has joined #openstack-tc23:39
*** kumarmn has quit IRC23:43
*** hongbin has quit IRC23:52

Generated by 2.15.3 by Marius Gedminas - find it at!