18:01:02 #startmeeting 18:01:02 Meeting started Thu Sep 16 18:01:02 2021 UTC. The chair is paddatrapper_. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:01:02 Useful Commands: #action #agreed #help #info #idea #link #topic. 18:01:33 agenda: https://deb.li/oNCD 18:01:42 #topic Roll call 18:01:49 Say hi if you're here for the meeting 18:01:51 Hi 18:02:57 Shall we wait for some more people or dive in? 18:03:32 can do. Otherwise we can motor through the agenda, but it'll be hard to reach consensus 18:04:00 yeah, and feedback around DC20 would be better with more people 18:04:57 highvoltage, wouter, olasd: you around? 18:05:06 (hi if you're here for the meeting) 18:05:12 (but not really) 18:05:29 hello I just arrived 18:06:13 #topic DC21 post-mortem 18:07:03 I guess what went right? what went wrong? 18:07:12 * pollo is lurking, but currently doing some other work-related stuff 18:07:25 a/v sync wasn't great 18:07:31 I think we should have had a policy of restarting voctomix daily 18:07:57 not sure I have anything significant to add after our last DebConf meeting 18:08:14 Any way of addressing the desync completely? 18:08:35 I wonder if we could detect it automatically 18:08:41 (in voctocore) 18:08:57 or in the ingests 18:09:22 voctomix2 (if that's still a thing?) is also a potential answer there 18:09:57 pre-review wasn't great. Probably because sreview wasn't ready to receive them until quite late 18:10:31 but also I think we probably should have trained the volunteers for that more, to get consistent review 18:10:50 yeah, we need to do more review training 18:10:55 both pre and post 18:11:09 during the event, we kept our heads above water, but we weren't as on top of upcoming talks as we should have been 18:11:35 We definitely lacked volunteers, especially later in the conf 18:11:38 it would have been good to be prepared for 1 or 2 talks in advance, recruiting volunteers, finding pre-recorded videos, etc. 18:11:47 but yeah, just not enough volunteer force 18:11:58 maybe better tooling there could have helped, too 18:12:10 (e.g. feeding pre-recorded video state back to wafer) 18:12:56 it would be nice to have 1 place to see the state of volunteer and source material for a talk 18:13:02 yeah 18:13:16 we could have done that 18:13:57 highvoltage: how painful was it to have 2 OBSs? 18:14:02 I also want to clean up and ansible the sreview downloader for the pre-recorded talks. Made it pretty easy to get all the talks for the day in the room 18:14:09 tumbleweed: it was... quite convoluted 18:14:37 tumbleweed: it has a concept of profiles, but it seemed like it didn't quite apply to everything (like extensions) 18:14:39 the point of it was to be able to have each one monitor voctomix state for its room, but we never really used that 18:14:45 (IIRC) 18:15:23 so, what happened was, if you would add more content to the first obs, and then close the first obs and then the second obs, the second obs would overwrite the config on exit with what it had 18:15:36 right 18:15:57 so it took some keeping wits about me that I close OBS's in the right order when closing them, and add content to the right one 18:16:57 which was also kind of tricky because you couldn't see the stream URI while the stream was running, and the second one often ended up being first in the window list which added to the confusion 18:17:37 it was a bit messy but managable. at least OBS didn't really have much problems other than that with two instances streaming to different places 18:18:16 sorry, was commuting 18:18:28 OBS still seems like the best tool for the job for fancy loops 18:19:07 it's also getting better, the advanced-scene-switcher added some new features that I requested during DC20, which was really nice 18:19:17 cool 18:19:36 and there are some other transition tools that I saw some ITPs for recently that will also fix a whole range of smaller issues 18:20:43 Anything else? 18:21:03 in terms of obs being best tool, there's probably ways we can make alternatives work 18:21:22 like, we could probably do most of OBS's tasks in a web page with some javascript if neede 18:21:35 yeah 18:21:46 I could imagine something like that running on the voctomix machine in every room at a debconf 18:21:49 I saw a nice tool from igalia too that could work (also with some significant elbow grease) 18:22:06 I think we won't be getting much better at recruiting volunteers for online confs 18:22:28 yeah, we need a bigger audience to get more volunteers 18:22:38 (hmm can't find the link for that now) 18:22:40 and constant adverts + daily training 18:23:02 we don't have the video-team t-shirt carrot 18:23:07 (although we could...) 18:24:17 o/ 18:24:25 sorry, almost forgot about the meeting 18:24:51 (my comment was more of the "online conference fatigue" department) 18:25:15 olasd: that too 18:26:09 i.e. I'm not sure we'll get even core volunteers enthusiastic about a 7-day online event again any time soon 18:26:47 :P 18:27:05 it kinda feels like it happened because "it had to", in a way 18:27:10 totally 18:27:31 imho for next year it might be nice if it has a very different format if online 18:27:50 olasd: yeah, agree with the "it had to" comment 18:28:07 highvoltage: do you have some ideas for what different format? 18:28:46 paddatrapper_: would take some brainstorming, one different idea was the fique em case use debian event that had one talk every evening for a whole month 18:29:17 yeah, the "ceph month" format had a half-day block every week for a month 18:29:23 paddatrapper_: but there's probably a whole range of things that could be done different 18:29:26 highvoltage: that is definitely an interesting format, but has a downside of "make it take much longer" 18:29:41 and the "every evening" bit is difficult for a world-wide event 18:29:45 wouter: indeed, it's just one way things have been different in the past, it doesn't need to be a month long 18:30:02 wouter: again, just an example of an online event that was different 18:30:33 highvoltage: sure, I got that the first time :) just pointing out the obvious issues that we'd have with such an option, but not at all saying we should therefore rule it out or anything 18:31:22 wouter: we don't /have/ to emulate an in-person DebConf format in an online event, there's probably a lot that could be done to deal with specific interests, specific languages, how we deal with different formats (like BoFs, talks, social chats, etc) 18:31:25 anyway I don't have anything to add on the subject. Maybe an "infra teardown" subtopic would be in order? 18:31:52 #topic DC21 post-mortem - Shut down remaining infra? 18:31:54 highvoltage: oh, absolutely! I think trying to emulate an in-person debconf is a bad idea to begin with, actually 18:32:02 wouter: yeah although I wasn't suggesting such an option either and I wouldn't either 18:32:37 so, what do we want to back up before shutting down infomaniak infra? 18:32:48 raw recordings need to be copied to vittoria 18:32:51 we've still got voctomix, obs, grabbers, encoders, prometheus 18:33:06 #info raw recordings need to be copied to vittoria 18:33:07 we haven't done that yet, I'll look at it tonight 18:33:08 we need to archive the pad contents 18:33:14 (that's on a grabber) 18:33:27 #action wouter to copy raw recordings to vittoria 18:33:29 #info we need to archive the pads content 18:33:35 (so I don't forget ;-) ) 18:33:40 I'll shut down and shelve the voctomixes 18:33:49 we only need them if we loose the raw video on encoder1 18:33:54 #action tumbleweed to shutdown and shelve the voctomixes 18:34:02 I'll also shutdown encoder2 18:34:09 highvoltage: and obs? 18:34:15 yeah, we can lose that one, as long as you don't touch encoder1 18:34:19 #undo 18:34:19 Removing item from minutes: 18:34:43 tumbleweed: it can just be destroyed 18:34:50 OK 18:34:59 #action tumbleweed to shutdown and shelve the voctomixes, encorder2, obs2 (just destroy) 18:35:18 I have a local backup of it, but there's likely nothing useful in there anyway 18:35:48 I guess grabber1 can go down too 18:35:51 pad is on grabber2 18:36:32 do we want anything from the prometheus node? 18:36:47 exporting metrics looks like a large PITA 18:36:48 The final machine sizes? 18:37:05 paddatrapper_: those are captured on a ticket 18:37:07 I'd say take a few screenshots of the CPU usage during the conf 18:37:16 so we can decide how to scale for next year 18:37:24 tumbleweed: cool. We didn't bump up the size of any of them? 18:37:25 but beyond that, I don't think we need to keep much? 18:37:48 https://salsa.debian.org/debconf-team/public/data/dc21-online/-/issues/1#note_256494 18:37:55 paddatrapper_: captured at the end of the event 18:38:00 great 18:38:17 #info our final stack: https://salsa.debian.org/debconf-team/public/data/dc21-online/-/issues/1#note_256494 18:39:51 anything else we need to shutdown/clean up? 18:40:35 think that's it 18:41:01 #topic Upgrade Hetzner infra to bullseye 18:41:09 shall we JFDI? 18:41:21 I guess the etherpad will have to be done manually for the moment 18:41:28 I think so? 18:41:30 sure 18:41:34 we know it works :) 18:41:35 it potentially can be used to fix the ansible role 18:41:40 yeah 18:41:48 although it's been 3 months and we haven't fixed the ansible role 18:41:59 I think we're just hoping that the next release of etherpad works better :P 18:42:05 (or we switch to hackmd / something) 18:42:14 I've probably done over a 100 different bullseye upgrades and it's probably been one of the most uneventful debian upgrades so far. so +1 for jfdi and if something doesn't work out, at least there's no conferences running right now 18:42:24 well there's also the vagrant thing that should help develop the ansible role 18:42:39 highvoltage: +1 18:42:59 any volunteers? 18:43:12 I may look at it in the next few days 18:43:16 (rain expected here) 18:43:36 #action tumbleweed to try get Hetzner infra upgraded to Bullseye 18:43:53 #topic MiniDebConf Regensburg - Hardware purchases 18:44:24 The budget for the long awaited purchases has been approved. Most of them are incoming 18:44:48 for instance, UPS is sending me 75 kg of gear from Thomann 18:45:08 #info the budget for the hardware purchases has been approved and things are on their way to olasd 18:45:18 75kg is... significant :) 18:45:42 yeah ouch! 18:46:00 olasd: you building your own cases? 18:46:09 tumbleweed: yeah 18:46:17 we look forward to pictures :) 18:46:30 it'll be a busy week :P 18:47:12 #topic MiniDebConf Regensburg - Who's there? 18:47:25 AFAIK it's tobi and me 18:47:28 I think it will be complicated for anyone from .za 18:48:46 * olasd looks at the registration page: wow, 50 people are registered 18:49:00 #info olasd and tobi will be there 18:49:07 (Belgian consulate told me that I'm not allowed in without a European vaccination passport, which is impossible to get in .za, even if I am fully vaccinated) 18:50:11 #topic Any other business 18:50:48 Just an FYI that I'm refactoring the ffmpeg commandline generation part of SReview into a separate library 18:51:06 (which it almost is, currently, just a few places where the integration is too close) 18:51:06 #info wouter is refactoring the ffmpeg commandline generation part of SReview into a separate library 18:51:26 should make it easier to handle video outside of SReview for future endeavors 18:51:38 e.g., the audio normalization that highvoltage wanted to do 18:51:39 olasd: I assume you'll need sreview for Regensburg? 18:51:55 that would be so nice 18:51:56 tumbleweed: if there's talks, yes, that's plausible 18:52:18 olasd: if there's a schedule, preferably in some pentabarf XML-like format, I can add it to SReview 18:52:35 wouter: I haven't seen a schedule yet. I assume it'll be a table on the wiki 18:52:36 olasd: (if it's not pentabarf XML I'll have to bang it into the db manually, which is doable but not great) 18:53:14 I'll let you know / have tobi let you know once there's a schedule 18:53:24 wfm 18:53:44 #topic Next Meeting 18:53:46 brownie points if it's easily parseable, but meh if not 18:53:59 I think we can reduce the frequency somewhat now, 18:54:03 s/,/?/ 18:54:03 Back to monthly? 18:54:23 sure? 18:54:33 when is the regensburg miniconf though? 18:54:45 monthly wfm 18:55:11 two weeks from now (29.09 - 03.10) 18:55:19 early next month apparently 18:55:28 I don't think we need another meeting before then, do we? 18:55:49 no 18:55:52 tobi is very quiet. 18:55:58 monthly wfm then, too 18:56:19 Back to every 3rd Thursday? (I think that what it was) 18:56:35 someone said my name... 18:56:51 I was busy spending Sponsor money :) 18:57:01 I was checking if something bad would happen if I say it three times in a row 18:57:16 21 October 2021 @ 18:00 UTC? 18:57:24 sgrm 18:57:26 sgtm* 18:57:31 paddatrapper_: +1 and thanks for keeping the wheels turning 18:57:52 #agreed 21 October 2021 @ 18:00 UTC? 18:57:58 highvoltage: heh, no worries 18:58:00 +1 18:58:01 without question mark I think ;-P 18:58:03 #undo 18:58:03 Removing item from minutes: 18:58:12 #agreed 21 October 2021 @ 18:00 UTC 18:58:15 lol 18:58:16 #endmeeting