| *** kumarmn_ has joined #openstack-tc | 00:43 | |
| *** kumarmn_ has quit IRC | 00:45 | |
| *** kumarmn has quit IRC | 00:45 | |
| *** kumarmn has joined #openstack-tc | 00:47 | |
| *** harlowja has quit IRC | 00:49 | |
| *** kumarmn has quit IRC | 00:50 | |
| openstackgerrit | Adrian Turjak proposed openstack/governance master: Adjutant official project status https://review.openstack.org/553643 | 00:52 |
|---|---|---|
| openstackgerrit | Adrian Turjak proposed openstack/governance master: Adjutant official project status https://review.openstack.org/553643 | 01:01 |
| *** david-lyle has joined #openstack-tc | 01:02 | |
| *** dtruong_ has joined #openstack-tc | 01:14 | |
| *** dtruong has quit IRC | 01:17 | |
| *** kumarmn has joined #openstack-tc | 01:37 | |
| *** kumarmn has quit IRC | 01:42 | |
| *** rosmaita has quit IRC | 01:46 | |
| *** kumarmn has joined #openstack-tc | 02:00 | |
| *** kumarmn has quit IRC | 02:04 | |
| *** kumarmn has joined #openstack-tc | 02:10 | |
| *** kumarmn has quit IRC | 02:14 | |
| *** mriedem has quit IRC | 02:26 | |
| persia | cdent: On IR toxicity: I argue it is in the way people use the tech, not the tech itself. Simply telling people not to /query tends to help a lot (I tend only to read /query if I know the person, and not with much priority). An assumption of multiplexors helps people appreciate the idea of "intermittent relay chat" (I cannot find the blog post that taught me that phrase), but I know several people who mostly deal with backscroll from public | 02:59 |
| persia | logs. | 02:59 |
| persia | Conversely, I've encountered communities where the normal length of an email is less than 100 characters, and the expected response is "now", to the point where setting clients to notify in 5-minute intervals causes people to be annoyed. | 03:00 |
| persia | So, rather than framing this as IRC vs. email, it might be useful to frame it in terms of expectation of immediacy. For a globally distributed community, immediacy is not achievable (at least not without either burning everyone out, or excluding anyone with a UTC offset of more than about 5 hours), regardless of the technology. | 03:01 |
| persia | Similarly, email can be sent to lists, or to individuals. Sending things to individuals puts undue burden on the individual, whether /query or email, and should be discouraged. | 03:02 |
| persia | I wish you the best of luck with your efforts to remove toxic behaviours from our communication, but I do encourage you to separate it from the technology employed (if only selfishly, as I have never found a mail client that doesn't annoy me when I am forced to use it, and generally prefer to read mailing lists as archives, usually subscribing only if archives aren't public, subscription lists are used for other purposes, or if one must be | 03:03 |
| persia | subscribed to post). | 03:03 |
| dmsimard | persia: I think it's all about managing expectations. If I give an example.. I don't answer phone calls if it's not from a number I recognize. And, people that know me also know that if it's important, to follow up with a SMS or Email (please no voicemail, seriously, if it's a voicemail it better be damn important). | 03:26 |
| dmsimard | You could say that appending |lunch or |afk to your IRC nickname manages people's expectations that you're probably not going to reply right away :D | 03:27 |
| dmsimard | But there's other ways to manage expectations, some of it is just built reputation (oh, this person always takes 2 hours to reply every time -- or this other one probably never sleeps) | 03:28 |
| *** kumarmn has joined #openstack-tc | 03:40 | |
| *** kumarmn has quit IRC | 03:58 | |
| *** harlowja has joined #openstack-tc | 04:35 | |
| *** kumarmn has joined #openstack-tc | 04:51 | |
| *** kumarmn has quit IRC | 04:55 | |
| *** harlowja has quit IRC | 05:22 | |
| *** cdent has joined #openstack-tc | 08:27 | |
| *** chandankumar is now known as chkumar|off | 08:42 | |
| *** kumarmn has joined #openstack-tc | 08:52 | |
| *** kumarmn has quit IRC | 08:56 | |
| *** jpich has joined #openstack-tc | 09:01 | |
| *** dtantsur|afk is now known as dtantsur | 09:42 | |
| ttx | EmilienM: flaper87: please rereview https://review.openstack.org/#/c/503145/ to remove or maintain your objection there | 10:16 |
| EmilienM | sure | 10:18 |
| ttx | https://wiki.openstack.org/wiki/Technical_Committee_Tracker updated, working on email now | 10:38 |
| ttx | Status email sent: http://lists.openstack.org/pipermail/openstack-dev/2018-March/128434.html | 11:05 |
| dims | thanks ttx | 11:10 |
| *** tbarron has joined #openstack-tc | 11:11 | |
| *** cdent has quit IRC | 11:54 | |
| *** rosmaita has joined #openstack-tc | 12:20 | |
| persia | dmsimard: Yes, but I think the key is to set reasonable expectations as a community, rather than just as individuals. Also, I find it frustrating when nicks change, but that may be related to client limitations (clients with automation to track nick changes make me less frustrated with them). | 12:28 |
| *** amonkV3WD5E has joined #openstack-tc | 12:40 | |
| *** mriedem has joined #openstack-tc | 13:06 | |
| fungi | well, it also presents issues for keeping track of users identified to registered nicks vs potential imposters | 13:19 |
| *** cdent has joined #openstack-tc | 13:22 | |
| openstackgerrit | Chris Dent proposed openstack/governance master: Fix stray ',' in trademark test location resolutiono https://review.openstack.org/553791 | 13:37 |
| openstackgerrit | Chris Dent proposed openstack/governance master: Fix stray ',' in trademark test location resolution https://review.openstack.org/553791 | 13:37 |
| persia | Indeed. Mediawiki does that by tying gerrit IDs to registered freenode accounts, but I think we've grown too big to adopt that model now. (it also presumes users register all the |foo nicks to their account) | 13:43 |
| *** kumarmn has joined #openstack-tc | 13:45 | |
| *** david-lyle has quit IRC | 13:45 | |
| cdent | smcginnis: ikr? | 13:48 |
| cdent | but I have a reputation to maintain here: I will type no word correctly before its time | 13:49 |
| smcginnis | ;) | 13:55 |
| cdent | Anybody have experience with adjutant? The history doc is very interesting, and the acknowledgement of the vagueness is good, but I'm struggling to get it | 13:57 |
| cdent | jroll: mugsie is the real author. I'm the ... editor? | 13:58 |
| smcginnis | I actually hadn't been aware of adjutant until the proposal. | 14:00 |
| cdent | same | 14:00 |
| *** hongbin has joined #openstack-tc | 14:00 | |
| jroll | cdent: you're right | 14:01 |
| * jroll finds more coffee | 14:01 | |
| cdent | me too please | 14:01 |
| jroll | :) | 14:03 |
| *** chkumar|off is now known as chandankumar | 14:07 | |
| mugsie | and, in the spirit of transparency, jroll is the original concept creator :) | 14:19 |
| jroll | lol. | 14:19 |
| jroll | all I did is ask if anyone discussed the option, and then ran away | 14:19 |
| jroll | don't credit me :) | 14:19 |
| mugsie | adjutant looks interesting, I could see that being *really* useful for large private cloud or public clouds | 14:19 |
| jroll | ++ | 14:20 |
| mugsie | We should record jroll's involvement for eternity, so we have someone to blame in 2 years ;) | 14:20 |
| * jroll thinks of at least one internal thing that adjutant could kill | 14:20 | |
| *** hongbin has quit IRC | 14:20 | |
| *** hongbin has joined #openstack-tc | 14:20 | |
| mugsie | jroll: same | 14:21 |
| cdent | mugsie: yeah, that's how adjutant looked to me too but it's hard to actually "see" it | 14:21 |
| *** dansmith is now known as superdan | 14:22 | |
| mugsie | Yeah, I think I am relying on a description I heard in a public cloud what session in Boston | 14:23 |
| tbarron | I have a concern w.r.t. Summit talk selection and | 14:29 |
| tbarron | the four opens: | 14:29 |
| tbarron | https://governance.openstack.org/tc/reference/opens.html | 14:29 |
| tbarron | and want to check if this channel is an appropriate forum for it. | 14:29 |
| cdent | tbarron: it may be, but be aware that summit talk selection is outside of the governance that the technical committee does. but somebody here might be able to provide some input or point you in the right direction | 14:30 |
| tbarron | cdent: ack, then I'll go ahead, understanding that this may not be in TC governance per se | 14:31 |
| dims | tbarron : yes, please go ahead and share. | 14:31 |
| tbarron | I would be +1000 for | 14:31 |
| tbarron | https://www.openstack.org/summit/vancouver-2018/summit-schedule/events/21040/building-nfs-service-based-on-manila-and-glusterfs-in-public-cloud | 14:31 |
| tbarron | except that it appears to violate all of the Four Opens. | 14:31 |
| tbarron | So far as I can tell, the work being presented is based on a | 14:32 |
| tbarron | fork of manila, not manila itself. You can't download or to my knowledge even | 14:32 |
| tbarron | buy the solution being presented in the talk. | 14:32 |
| tbarron | Upstream we need maintainers for the glusterfs | 14:32 |
| tbarron | back ends (non-voting job in CI gate is failing, again) but no one has | 14:32 |
| tbarron | contributed to maintenance of that back end for several years. | 14:32 |
| tbarron | Upstream we have active discussions of the SPOF issues that they claim | 14:32 |
| tbarron | to have solved - on irc, in spec proposals, in the weekly meeting, | 14:32 |
| tbarron | on emails, and at the PTG. These guys may be listening to this discussion | 14:32 |
| tbarron | but they aren't contributing to it. | 14:32 |
| cdent | ah | 14:33 |
| tbarron | That's it - btw I am asking here in my (new) manila PTL capacity. | 14:33 |
| tbarron | I know there's nothing to stop forks of openstack, my issue is with us helping promote these. | 14:33 |
| cdent | dims you were at the board meeting at the ptg, yes? wasn't there some stuff related to lack of upstream contribution there? | 14:33 |
| cdent | tbarron: yeah, definitely problematic | 14:34 |
| tbarron | Ideal outcome would be a way to integrate this develpment/innovation back into core upstream work. | 14:34 |
| * cdent nods | 14:34 | |
| tbarron | Also, that selection process for talks be aware of such issues, my guess is that everyone thought the | 14:34 |
| tbarron | work presented was really manila itself, not a fork. | 14:35 |
| tbarron | Theoretically maybe their innovation is 100% outside of manila proper, in some deployment wrapper. But | 14:35 |
| dims | tbarron : totally fair that work needs to find way back to core upstream. | 14:35 |
| tbarron | that seems doubtful and in any case they aren't contributing to gluster backend/driver code maintenance in manila. | 14:36 |
| jroll | I've seen many talks like this, fwiw, I just always crossed my fingers and hoped the work would make its way upstream eventually | 14:36 |
| persia | While I sympathise with the particular example, my general experience is that many talks at summits past have shared the "this is the new thing we are doing, which we're intending to maybe send upstream" spirit, and nearly everything on the expo floor ranges from that to "this is our proprietary extension for $service". | 14:36 |
| persia | I always felt the talks were accepted for funding reasons, and considered them an opportunity to discuss closer collaboration, upstreaming, etc. | 14:37 |
| jroll | then again, I've seen shuttleworth give a keynote about MaaS being better than ironic, so I don't really assume speakers care about openstack upstream | 14:37 |
| dims | tbarron : there are a couple of things we can do ... there are track chairs for the various tracks, we can raise awareness. also the cores from manila can go to the talk and engage with these folks | 14:37 |
| cdent | tbarron: besides anything else that might happen I would contact the speakers and make sure they know you'd like to see if it is possible for the code to come upstream. We've seen plenty of people, who, for whatever reason, had no idea. | 14:38 |
| tbarron | dims: I am planning on going to the talk and actually want to seek a way to disclose the questions I would ask to the speakers ahead of time. | 14:38 |
| jroll | persia: I agree with everything you said | 14:38 |
| dims | tbarron : let me see if we can find a way to reach them proactively | 14:38 |
| tbarron | I want to give them a chance to say "we're going to do the right thing" or to announce that pro-actively so I don't even have to say anything but express apprecciation. | 14:38 |
| tbarron | with one 'c' instead of two | 14:39 |
| cdent | tbarron: are you feeling like these people know what they are doing and taking advantage, or is there a chance they are just doing what they think is right but are ill informed? | 14:41 |
| dims | cdent : i'd err on the side of they are new, lets try to engage them | 14:43 |
| smcginnis | When I was a track chair I would usually try to filter out anything that looked like a product pitch. | 14:43 |
| cdent | dims: that's what I'm hoping | 14:43 |
| smcginnis | Or the ones that outright said "if this talk gets accepted we will open source our code". | 14:43 |
| smcginnis | This looks like it would not have been clear to the chairs unless they had a deeper understanding of the state of things within Manila. | 14:44 |
| dims | tbarron : i've dm'ed their email id(s) to you, wanna start a thread? | 14:44 |
| dims | smcginnis : right, looking at the list of track chairs for Public Cloud, don't think they were deep into Manila | 14:44 |
| *** annabelleB has joined #openstack-tc | 14:44 | |
| tbarron | cdent: I don't assume malice. | 14:45 |
| tbarron | smcginnis: right, but maybe there should be some check-boxes for talk submissions, or guidance on this front at least? | 14:46 |
| tbarron | smcginnis: we can't get perfection in the process but we can influence it positively I think. | 14:46 |
| smcginnis | ++ | 14:46 |
| dims | tbarron : i am more inclined towards, folks like this exist, let's use the platform we have to get them into the community and work with us. | 14:47 |
| tbarron | dims: +1000 | 14:47 |
| smcginnis | When I had done it, there was some talk of documenting some guidelines, but at least at the time it was pretty freeform and up to each track how they wanted to evaluate and handle selections. | 14:47 |
| smcginnis | dims: That would be ideal. | 14:47 |
| dims | smcginnis : as part of the track process there is a way to reach potential presenters and evangelize things as well | 14:48 |
| smcginnis | ++ | 14:49 |
| dims | smcginnis : for example if that talk had not gotten accepted then we would not have had the chance for tbarron to find these potential new contributors | 14:49 |
| dims | tbarron : smcginnis : we can get folks from SIG-firstcontact to go talk to them as well at the summit to figure out how to increase contributions from China Mobile | 14:51 |
| *** chandankumar is now known as chkumar|off | 14:52 | |
| tbarron | Yeah, I like the SIG-firstcontact idea. | 14:55 |
| dims | diablo_rojo : ^ | 14:57 |
| dims | diablo_rojo : we could use some help with talking to china mobile folks working on manila | 14:58 |
| * tbarron passes diablo_rojo a cup of coffee, 2nd day in a row | 15:00 | |
| dims | :) | 15:00 |
| tbarron | not that Seattle's Best stuff either | 15:01 |
| *** melwitt is now known as jgwentworth | 15:08 | |
| openstackgerrit | Jeffrey Zhang proposed openstack/governance master: Move kolla-kubernetes deliverable under governance of openstack tc https://review.openstack.org/552531 | 15:10 |
| *** ianychoi_ has joined #openstack-tc | 15:53 | |
| *** ianychoi has quit IRC | 16:03 | |
| *** knikolla has quit IRC | 16:03 | |
| *** harlowja has joined #openstack-tc | 16:03 | |
| *** dtruong_ has quit IRC | 16:04 | |
| *** dtruong has joined #openstack-tc | 16:05 | |
| *** knikolla has joined #openstack-tc | 16:11 | |
| *** dtantsur is now known as dtantsur|afk | 16:31 | |
| *** annabelleB has quit IRC | 16:33 | |
| *** annabelleB has joined #openstack-tc | 16:42 | |
| *** annabelleB has quit IRC | 16:55 | |
| *** annabelleB has joined #openstack-tc | 17:01 | |
| *** harlowja has quit IRC | 17:05 | |
| *** kumarmn has quit IRC | 17:06 | |
| *** diablo_rojo has quit IRC | 17:14 | |
| *** diablo_rojo has joined #openstack-tc | 17:14 | |
| *** pabelanger has quit IRC | 17:17 | |
| *** pabelanger has joined #openstack-tc | 17:17 | |
| *** kumarmn has joined #openstack-tc | 17:19 | |
| *** kumarmn has quit IRC | 17:23 | |
| *** ianychoi__ has joined #openstack-tc | 17:24 | |
| *** ianychoi_ has quit IRC | 17:27 | |
| *** jpich has quit IRC | 17:32 | |
| *** annabelleB has quit IRC | 17:57 | |
| *** annabelleB has joined #openstack-tc | 18:01 | |
| *** harlowja has joined #openstack-tc | 18:01 | |
| *** harlowja_ has joined #openstack-tc | 18:04 | |
| *** harlowja has quit IRC | 18:06 | |
| diablo_rojo | tbarron, dims smcginnis, just caught up | 18:17 |
| tbarron | diablo_rojo: good morning :) | 18:18 |
| diablo_rojo | My client disconnected and reconnected and lost the ping apparently | 18:18 |
| diablo_rojo | I promise I've been up for more than an hour :) | 18:18 |
| tbarron | diablo_rojo: just messing with you of course | 18:18 |
| * diablo_rojo reads through the abstract quickly again | 18:18 | |
| tbarron | diablo_rojo: I'm going to wait at least a day to send any email; want to be completely diplomatic, constructive. | 18:19 |
| tbarron | diablo_rojo: goal here is to get them to contribute upstream and if they are using glusterfs help maintain that driver, not to shame or blame. | 18:19 |
| persia | timezones are flexible. It is as early as 7:20 Friday in some parts of the world. | 18:19 |
| diablo_rojo | tbarron, and thats why we like you :) | 18:19 |
| * diablo_rojo goes to look at the speakers involvement in things- gerrit hist etc | 18:20 | |
| tbarron | diablo_rojo: at the same time, if there is no desire to contribute upstream, then I'd like the selection process to start taking that kind of thing into account. | 18:20 |
| diablo_rojo | tbarron, I would be inclined to agree | 18:22 |
| diablo_rojo | Looks like the first speaker has done a lot in karbor and in the manilaclient, but that's fine I think | 18:23 |
| diablo_rojo | *so thats fine | 18:23 |
| diablo_rojo | The second one I can't find in gerrit and the last one..has only touched Nova and not since 2016 | 18:25 |
| persia | Tracking affiliation changes may be useful as well, as not all individuals who have contributed in the past may continue to do so (and not all who were pure downstream in the past stay that way if permitted to work upstream) | 18:27 |
| diablo_rojo | tbarron, so the action is to email them and say..? | 18:27 |
| tbarron | diablo_rojo: wow, looks like you folks are doing great stuff with manila and glusterfs. | 18:29 |
| tbarron | diablo_rojo: do you have plans to contribute it to upstream manila and help support the glusterfs driver? | 18:29 |
| diablo_rojo | persia, yeah I started looking at member profiles too- Hecheng doesn't have one- the other two do (Pengju acutally has three it looks like) so thats promising. | 18:29 |
| tbarron | diablo_rojo: you are working in an area we've been discussing and we could really use support for that driver from people who kknow it and use it. | 18:30 |
| tbarron | diablo_rojo: somethiing like that. | 18:30 |
| tbarron | diablo_rojo: but I want to wait, get ideas, etc. first. | 18:30 |
| dhellmann | it seems quite reasonable to me to email and say that you would welcome their contributions to manila and to point out that you especially need help with the glusterfs driver and then to ask what sorts of things we can do to help them contribute the work they are doing | 18:31 |
| diablo_rojo | tbarron, that seems like a good approach. Good idea to wait a bit too. | 18:31 |
| diablo_rojo | dhellmann, +1 | 18:31 |
| dhellmann | which is not quite how you had it phrased, tbarron, but I think reflects the intent you had | 18:32 |
| tbarron | that is the intent | 18:32 |
| dhellmann | "we would like to work with you" feels a little different from "are you going to work with us" | 18:32 |
| tbarron | Agreed. | 18:33 |
| *** kumarmn has joined #openstack-tc | 18:33 | |
| diablo_rojo | If they need help getting started upstream I would happily do so- there is the training that we will be doing right before the summit so maybe they can stop by there and I can help? | 18:33 |
| dhellmann | good idea | 18:33 |
| tbarron | dhellmann: at the same time I don't want anyone surprised if they do not decide to engage if I go to the talk | 18:33 |
| tbarron | and publicly ask if they intend to contribute any of this stuff that they are presenting upstream | 18:34 |
| dhellmann | sure. I think that's my point though. "confronting" them by asking if they will contribute may give different results than simply "inviting" and "encouraging" them to do so | 18:35 |
| tbarron | dhellmann: so I want to have this "invitation" out early, and if there are wiser people than I w.r.t. the possible followup then | 18:35 |
| tbarron | I want their counsel too. | 18:35 |
| dhellmann | you'll catch more flies with honey, etc. | 18:35 |
| tbarron | dhellmann: yup, let me draft an email by early next week and run it by diablo_rojo dhellmann dims smcginnis et. al. before I send it | 18:36 |
| tbarron | really anyone interested in consulting on it | 18:37 |
| dhellmann | tbarron : you're probably best placed to help them directly with any contributions, but I'm sure all of us would be happy to help facilitate any conversations if you'd like input | 18:37 |
| dhellmann | tbarron : sure, I'll gladly review a draft if you like | 18:39 |
| *** kumarmn has quit IRC | 18:40 | |
| tbarron | diablo_rojo: dhellmann ty | 18:46 |
| diablo_rojo | tbarron no problem :) Happy to review. | 18:47 |
| dhellmann | tbarron : thank you for trying to take this on in a constructive way :-) | 18:50 |
| *** cdent has quit IRC | 18:51 | |
| *** kumarmn has joined #openstack-tc | 18:52 | |
| tbarron | persia: you are certainly right about affiliations changing and it could even be that I've talked to some of these folks on IRC, say, and don't know it. | 18:55 |
| *** annabelleB has quit IRC | 19:16 | |
| *** annabelleB has joined #openstack-tc | 19:20 | |
| *** annabelleB has quit IRC | 19:25 | |
| *** annabelleB has joined #openstack-tc | 19:36 | |
| openstackgerrit | Adrian Turjak proposed openstack/governance master: Adjutant official project status https://review.openstack.org/553643 | 19:59 |
| dims | thanks dhellmann tbarron diablo_rojo | 20:08 |
| *** annabelleB_ has joined #openstack-tc | 20:11 | |
| *** annabelleB has quit IRC | 20:13 | |
| *** annabelleB_ is now known as annabelleB | 20:13 | |
| *** annabelleB has quit IRC | 20:22 | |
| *** annabelleB has joined #openstack-tc | 20:39 | |
| *** annabelleB has quit IRC | 20:58 | |
| fungi | tbarron: i sympathize on the private manila fork talk... brings back memories of when a certain keynote presented metrics for a (up to that point at least) private rewrite of nova in go (even the conference organizers didn't learn the details of that one until moments before it was on stage) | 21:41 |
| fungi | not that a first-contact sig, if we'd had one back then, would have been much help to that group | 21:43 |
| smcginnis | fungi: That was the oddest, most confusing keynote I think I've seen. | 21:44 |
| smcginnis | Never saw so many attendees looking at each other confused and scratching their heads. | 21:45 |
| * mtreinish is glad he slept in that morning | 21:45 | |
| tbarron | fungi: did you succeed in bringing that fork back into the fold? | 21:45 |
| tbarron | guess not if it was essential to be in go | 21:45 |
| fungi | not so much | 21:45 |
| smcginnis | Apparently that's actually used by some folks, but I don't think there was any interest to have that merged back in. | 21:46 |
| clarkb | it also wasn't so much a fork as a rewrite the smallest number of features to get a big splash on stage | 21:46 |
| smcginnis | Exactly. | 21:47 |
| fungi | yeah, punted on the hard problems and then congratulated itself on how quickly you can solve the easy ones | 21:47 |
| tbarron | well let's see what we can do with this one; smcginnis and xyang and dims are going to help me open manila up to china | 21:47 |
| fungi | the eastern passage! | 21:47 |
| smcginnis | :) | 21:48 |
| tbarron | I'm being somewhat facetious but we have a lot of interest and people who want to contribute and we | 21:48 |
| tbarron | have challenges w.r.t. how to take our old ways of doing things, welcome new contributors, and | 21:48 |
| tbarron | maintain the valuable, etc. | 21:48 |
| fungi | i have to imagine chinese industry also needs shared filesystems, at least occasionally | 21:48 |
| tbarron | fungi: chinese industry, us financial enterprises, telcos are most interested atm; interesting mix | 21:49 |
| fungi | quite awesome | 21:49 |
| tbarron | telcos of course want storage everywhere including "edge" where there is no bandwidth back to anything, :) | 21:50 |
| tbarron | "just" classic caching issues | 21:51 |
| fungi | cache it, problem solved, right? ;) | 21:51 |
| openstackgerrit | Tony Breeds proposed openstack/project-team-guide master: Update Stable policy for Extended Maintenance https://review.openstack.org/552733 | 21:53 |
| tonyb | project ciao was never inteneted to be folded back in it was supposed to create a discussion about how we can improve OpenStack. The Keynote was a bad choice to bring it up. It shoudl have been a session in the design summit as the intended audience was devs not consumers | 22:03 |
| tonyb | It's still ticking along but as you add more bits that nova does it looses some of it's performance. | 22:03 |
| smcginnis | Imagine that. :) | 22:04 |
| tonyb | smcginnis: the discussion then moves to making other services (glance, keystone etc) perform better (by re-writing in go and droppin cruft) and then it starts to look really big | 22:05 |
| tonyb | It would have been soooo much cooler if there was a detailed performance analysis of what things are actually slow and then *why* they're slow | 22:06 |
| smcginnis | I do agree there is a lot of cruft, and we probably could improve performance in some cases by rewriting in something other than python. But I think the reality of backwards compatibility, upgrades, etc. are too easy to overlook when presented the way that was. | 22:06 |
| smcginnis | An actual analysis and some actionable data could have made that way, way better. | 22:07 |
| tonyb | I recall a while ago there were claims that designate (sp?) in python was always going to be too slow ... and then once the gauntlet was down in < 50lines of changes the speed was suddenly on par with go | 22:07 |
| *** knikolla has quit IRC | 22:07 | |
| *** hongbin has quit IRC | 22:07 | |
| smcginnis | I think I remember that. | 22:07 |
| *** knikolla has joined #openstack-tc | 22:07 | |
| *** hongbin has joined #openstack-tc | 22:07 | |
| tonyb | Yup. I kinda which I had resources / time to do that. Reduce the overall memory consumption and get a feel for performance | 22:08 |
| mriedem | missing some critical punctuation there tony | 22:08 |
| tonyb | soo many "fun" things to do | 22:08 |
| clarkb | many/most/maybe all openstack services are io bound | 22:08 |
| clarkb | python does reasonably well in those cases | 22:08 |
| clarkb | but just like it is possible to write slow C or Go you can write slow python | 22:09 |
| tonyb | clarkb: Sure. | 22:09 |
| smcginnis | I will not be surprised when we move to py3 only if we can find some ways to improve performance in some key areas. | 22:09 |
| tonyb | clarkb: I'm also not sure that "openstack image list" is I/O bound but it sure feels slow ... not picking on glance | 22:10 |
| clarkb | tonyb: it actually is. If you have a few minutes I can talk about it :) | 22:10 |
| clarkb | basically a large overhead in `openstack *` is pkg_resources loading and sorting all the findable python packages so that it can load entrypoints | 22:10 |
| clarkb | this scales in terms of your disk's performance and in number of python packages you have installed | 22:11 |
| clarkb | and it isn't linear beacuse of the sort | 22:11 |
| clarkb | all that before a single http request can be made | 22:11 |
| tonyb | Hmm okay I see where you're going. | 22:11 |
| clarkb | (I think on my local machine when I was debugging that about half the cost of running `openstack foo` was just pkg_resources the other half was network rtt for http) | 22:11 |
| clarkb | this would be a case of writing slow python (in the name of easy pluggability | 22:12 |
| clarkb | but python doesn't have to be slow like that | 22:12 |
| cmurphy | this was one of the analyses on designate performance http://lists.openstack.org/pipermail/openstack-dev/2016-May/094568.html | 22:13 |
| clarkb | if pkg_resources was in stdlib I'd blame python a bit more but it is part of setuptools instead (which is de facto stdlib) | 22:13 |
| tonyb | cmurphy: Thanks that | 22:15 |
| tonyb | s the one :) | 22:15 |
| tonyb | clarkb: Yeah okay so I was under thinking it ;P | 22:15 |
| * tonyb relocates | 22:15 | |
| clarkb | so did designate end up addressing those issues in python? | 22:17 |
| clarkb | if so cool | 22:17 |
| fungi | on rewriting, i bet lots of openstack could be made way faster by rewriting... in python... and dropping cruft | 22:26 |
| smcginnis | Indeed | 22:26 |
| fungi | question is who has time to rewrite enough of it in any language | 22:27 |
| persia | Or incentive. I've seen a few "let's rewrite OpenStack" efforts over the years (and in some ways OpenStack was a "let's rewrite foo" effort initially). Only one of them seems to remain operational, but has moved to a slightly different space (with a better mission goal) now.. | 22:29 |
| persia | I think it was more exciting for firms to launch "The OpenStack Replacement" at a different time in terms of marketing spend. Now, firms might end up replacing many layers of things, but would probably call it something else. | 22:30 |
| dims | fungi : the 5+ libvirt/kvm projects in k8s eco system are going to try :) | 22:31 |
| persia | (yes, "bare metal hosting for containers", "container virtualisation", "fully isolated container infrastructure" and similar are likely words, as those projects demonstrate) | 22:32 |
| mtreinish | fungi: edunham has a good talk on that topic: https://www.youtube.com/watch?v=6jqy-Dizd0I | 22:41 |
| *** annabelleB has joined #openstack-tc | 22:46 | |
| *** hongbin has quit IRC | 22:48 | |
| *** kumarmn has quit IRC | 22:48 | |
| *** superdan is now known as dansmith | 22:58 | |
| *** kumarmn has joined #openstack-tc | 23:05 | |
| *** kumarmn has quit IRC | 23:25 | |
| *** kumarmn has joined #openstack-tc | 23:30 | |
| *** openstackgerrit has quit IRC | 23:48 | |
| *** jgwentworth is now known as melwitt | 23:48 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!