[01:02:34] Analytics-Kanban, VisualEditor: Schema:Edit seems to incorrectly set users as anonymous {lion} - https://phabricator.wikimedia.org/T92596#1208537 (Krenair) a:Krenair>None [01:49:46] Analytics-EventLogging, MediaWiki-General-or-Unknown, Patch-For-Review, Performance: Add event tracking queue to MediaWiki core for loose coupling with EventLogging or other interested consumers - https://phabricator.wikimedia.org/T95356#1208569 (bd808) What should the config setup for the EventLog... [03:08:55] (PS1) Nuria: Undoing Ide98e20eb54523353153ccd212df5... [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/204214 (https://phabricator.wikimedia.org/T93023) [03:21:04] (PS2) Nuria: Undoing Ide98e20eb54523353153ccd212df511a9298bd16 [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/204214 (https://phabricator.wikimedia.org/T93023) [03:22:54] (CR) Nuria: [C: 2] Undoing Ide98e20eb54523353153ccd212df511a9298bd16 [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/204214 (https://phabricator.wikimedia.org/T93023) (owner: Nuria) [03:23:01] (Merged) jenkins-bot: Undoing Ide98e20eb54523353153ccd212df511a9298bd16 [analytics/wikimetrics] - https://gerrit.wikimedia.org/r/204214 (https://phabricator.wikimedia.org/T93023) (owner: Nuria) [04:08:39] Analytics-EventLogging, Analytics-Kanban: Flow events not validating on EL - https://phabricator.wikimedia.org/T95169#1208672 (Mattflaschen) In the future, just associating the #Flow project will ensure we see it. [04:10:13] Analytics-EventLogging, Analytics-Kanban: Flow events not validating on EL - https://phabricator.wikimedia.org/T95169#1208674 (Mattflaschen) But this is actually an #Echo bug. [04:10:24] Analytics-EventLogging, Analytics-Kanban, Collaboration-Team, Echo: Flow events not validating on EL - https://phabricator.wikimedia.org/T95169#1208675 (Mattflaschen) [04:11:35] Analytics-EventLogging, Analytics-Kanban, Collaboration-Team, Echo: Echo events not validating on EL - https://phabricator.wikimedia.org/T95169#1208678 (Mattflaschen) [04:19:53] Analytics-EventLogging, Analytics-Kanban, Collaboration-Team, Echo, Patch-For-Review: Echo events not validating on EL - https://phabricator.wikimedia.org/T95169#1208688 (Mattflaschen) a:Mattflaschen [08:01:52] Analytics-Cluster, Analytics-Kanban: Compute pageviews aggregates daily and monthly from April - https://phabricator.wikimedia.org/T96067#1208840 (JAllemandou) [08:02:22] Analytics-Cluster, Analytics-Kanban: Compute pageviews aggregates daily and monthly from April - https://phabricator.wikimedia.org/T96067#1207516 (JAllemandou) Thanks for the reminder Aklapper ! I usually add projects, and forgot this time :) [13:07:28] o/ joal [13:17:02] Today I am focusing 100% on moving value-adding/WikiCredit forward. :) [13:33:20] hey halfak [13:33:37] Sorry, I was concentrated :) [13:33:50] I have 1/2 hour to talk now if you wish :) [13:36:11] Oh yeah! Batcave? [13:36:14] joal, ^ [13:36:15] sure :) [15:04:39] joal, something is totally weird [15:04:40] e.g. [15:04:41] https://yarn.wikimedia.org/proxy/application_1424966181866_85838/mapreduce/job/job_1424966181866_85838 [15:04:48] that is the add partition job for a webrequest misc hour [15:04:50] very low data [15:04:54] it has been running for 20 hours [15:05:06] i can't find the actual job though, that is the oozie launcher task [15:05:16] i think it shoudl spawn off another application [15:05:39] AttemptID:attempt_1424966181866_85838_m_000000_0 Timed out after 600 secs Container released on a *lost* node [15:05:52] maybe this has something to do with analytics1017 crashing this weekend [15:06:39] going to kill and restart that one...via oozie [15:06:40] let's see [15:21:25] joal: i;m going to kill your SELECT regexp_extract(uri_host,...100000(Stage-1) job [15:21:27] it hasn't started yet [15:38:54] joal, simplewiki XML on altiscale HDFS @ /user/halfak/streaming/simplewiki-20141122/xml-bz2 [15:56:57] thx halfak [16:00:44] joal, you talked to yurik a bunch yesterday about his jobs, right? [16:00:45] they ahve been running a long time. [16:02:18] mforns_brb: let me know if you need help with deploying EL so you get your logging change [16:05:11] nuria, I'm going to look at the docs for deploying, and if I need something I ping you [16:05:20] thanks! [16:12:06] joal, yt? [16:12:15] mforns: k [16:14:19] lunch! [16:15:31] nuria, I do not have access to tin.eqiad.wmnet :[ [16:15:58] ok, i think i figured out what was wrong with the cluster, although I don't know exactly why it went wrong. [16:16:01] no wlunch! [16:16:43] ottomata, should I file a task for ops to request access to tin.eqiad.wmnet? [16:18:16] nuria, from the EL docs, it seems that I do not need root in tin.eqiad.wmnet to deploy EL. But I guess I do need root, right? [16:21:02] sorry ottomata didn't see your lunch, we talk later :] [16:22:17] o/ kevinator [16:22:21] WOops. offline. [16:22:24] * halfak --> email [16:26:16] ottomata: here now, but in meeting [16:26:23] Can we talk after ? [16:28:57] ottomata: I am eager to know what it is ! [16:30:02] mforns: i think so yes. [16:30:28] mforns: or maybe just ssh access, did you try that? [16:30:55] nuria, what do you mean? ssh tin? [16:31:00] yes [16:31:33] yes, that's what I tried [16:31:39] nuria, is there another waY? [16:32:03] mforns: so you DO have ssh, correct? [16:32:35] when I execute: ssh tin.eqiad.wmnet [16:32:46] nuria, I get: Permission denied (publickey) [16:33:02] mforns: ok, you need to file a request for access [16:33:10] mforns: i will deploy [16:33:19] nuria, yes, and ask for root [16:33:33] mforns: no actually w/o root it works [16:33:46] mforns: i just forget cause it is different everywhere [16:33:51] nuria, does it take a long time, because, maybe I can talk with andrew and speed up the access [16:33:59] nuria, ok [16:34:28] mforns: it's fine, it'll take 1 week but it is ok, cause really, as long as you have access to EL box you can change the code, right? [16:34:36] as we deploy from source here [16:35:45] nuria, I meant if the deployment process does take a long time? so that you don't need to do that [16:36:07] yurik: yt ? [16:36:20] joal, ? [16:36:41] As per ottomata email, please don't run jobs on the cluster now [16:36:55] I know that it has catch up, but still :) [16:37:03] joal, i'm not, there is a script that autoruns things [16:37:05] There is some debugging and config change needed [16:37:13] Then please stop this script :) [16:37:23] \it runs the thing on your name, you run it ;) [16:37:25] working on that already :))) [16:37:33] ok [16:37:38] eta? [16:37:45] Just to let you know: if needed, we'll kill themn [16:37:51] Thamks :) [16:37:55] when can i start them again? [16:38:02] asap, hoppefully [16:38:11] ok, need about 10 min [16:38:27] we'll let you know asap it's readcy [16:38:39] Thanks again :) [16:41:21] mforns: note that the master also includes a graphite change that you will need to foloow up on: [16:41:25] https://www.irccloud.com/pastebin/56wRNqhU [16:42:25] nuria, ok [16:44:21] mforns: ok, deployed to EL box, try and see if code made it to hafnium and you can deploy there [16:44:34] nuria, ok [16:45:51] mforns: please check EL is doing good with your changes [16:46:01] mforns: brb in 40 mins [16:46:08] nuria, it seems [16:46:21] I have no access to hafnium too... [16:46:37] sigh... sorry, I'll ask for it too [16:46:38] mforns: ok, another one [16:47:00] mforns: looks like EL is fine but do check logs for abit [16:47:08] ok [16:48:23] joal: back from lunch, lemme know when you want to chat [16:51:38] mforns: uh, yes, if you want deployment access ja [16:51:41] make a ticket [16:55:38] ottomata, thanks already doing that [17:38:48] Analytics, WMF-Product-Strategy: Backfill pageview data for March 2015 from sampled logs before transition to UDF-based reports as of April - https://phabricator.wikimedia.org/T96169#1209816 (DarTar) NEW [17:42:26] oo, nuria, joal, i set up a stupid nginx reverse proxy on my mac, now I don't have to replace yarn.wikimedia.org when clicking on links naymore [17:46:47] hey Ironholds! [17:46:56] you starting jobs/!? [17:47:53] o/ mili|lunch [17:48:02] When you get back, check out this: https://commons.wikimedia.org/wiki/File:Diff_time_density_(enwiki).svg [17:48:16] (PS5) Jdlrobson: Update limn graphs [analytics/limn-mobile-data] - https://gerrit.wikimedia.org/r/199162 (https://phabricator.wikimedia.org/T93690) [17:49:00] The median time it takes to diff article revisions in Wikipedia is 0.05 seconds. The 90th percentile is 0.23 seconds. [17:50:31] So, we're super-duper fast at performing diffs. I'm now looking back to hadoop for behaving badly with the reducer. [17:51:26] halfak: that's with spark? [17:51:56] ottomata, streaming mapreduce where I do most of the work in the mapper. [17:53:45] aye cool [17:54:32] ottomata, did I tell you that the big enwiki job finished? [17:54:37] no! [17:54:39] awesome [17:54:45] On our smaller altiscale cluster, I was able to complete it in 5 days [17:54:48] :) [17:55:03] So, now I'm doing analysis on it to figure out how long the bits took and how big the diffs are. [17:55:43] I don't have a good sense for out capacity at altiscale right now, but joal mentioned that he thought it was only 4 boxes. [17:56:32] Analytics-Kanban: Safely reboot limn1, wikimetrics*, dan-pentaho, and any other labs instances running Ubuntu Precise - https://phabricator.wikimedia.org/T96175#1209914 (Milimetric) NEW a:Milimetric [17:57:33] nuria: The first step on Developer setup for EL - where does this code go? [18:03:04] ottomata, hi, can i run my scripts again? [18:06:45] yurik: i believe i have fixed the main problem, which had to do with a weirdly stuck job, not lots of folks usin gthe cluster, so that is good. [18:07:03] however, i would like to verify this, and i would like to make sure refinement jobs get caught up asap before I say you can submit more jobs [18:07:11] ok [18:07:39] Analytics-Tech-community-metrics: Ensure that most basic Community Metrics are in place and how they are presented - https://phabricator.wikimedia.org/T94578#1167361 (Qgil) fyi, I was right about this task being a high priority goal for this quarter. ;) >>! In T94165#1209948, @Qgil wrote: > The main piece... [18:10:06] ottomata: I have time now :) [18:10:40] halfak: that's music to ears, great stuff [18:16:16] halfak: spark not really set up in altiscale from what I see [18:16:36] We can file a ticket for it. [18:16:38] I don't manage to get my job running, and spark-shell is not ready to use [18:16:42] It *should* be set up. [18:16:43] I did ;) [18:16:49] Great :) [18:17:09] Got help already about spark-shell (port between machines issue, should be fixed soon) [18:17:22] And just logged a new one about job failing without exploitable logs [18:18:17] joal, on phone... [18:18:32] no problemo ottomata, will wait for you ;) [18:20:26] milimetric: Around? [18:20:53] hi madhu [18:21:37] Analytics, Analytics-Kanban, WMF-Product-Strategy: Backfill pageview data for March 2015 from sampled logs before transition to UDF-based reports as of April - https://phabricator.wikimedia.org/T96169#1209980 (kevinator) [18:21:40] question on EL again. The first step on Developer setup for EL (https://www.mediawiki.org/w/index.php?title=Extension:EventLogging#Specifying_a_dependency_on_EL) - where does this code go? [18:22:05] also what does it do and is it necessary [18:22:28] madhuvishy: so this is developer as in developer using EL, not working on EL [18:22:35] but working on EL includes using EL [18:22:36] milimetric: aaah [18:23:17] but are you wondering more generally about EL? We can hangout and talk if you like [18:24:34] madhuvishy: EL is a service [18:24:46] So this was the first task on my onboarding - https://phabricator.wikimedia.org/T89162. I'm in general wonderment on where to start. [18:24:51] madhuvishy: in vagrant you only have access to client side part [18:25:07] madhuvishy: that logging goes on the server side [18:25:12] nuria: right. [18:25:21] madhuvishy: do you have the code chcked out? [18:25:26] *checked [18:25:38] yes [18:25:43] madhuvishy: and the role set up? [18:25:56] madhuvishy: ok, set up also navigation timing rile [18:25:57] *role [18:26:00] role set up in vagrant yes. [18:26:02] and we can go to batcave [18:26:25] ah okay, let me provision for that role [18:29:08] madhuvishy: also do you have permits to ssh to machines in labs? [18:29:47] mforns: did you check your login in EL box? [18:30:08] nuria: https://phabricator.wikimedia.org/T96053 says its done [18:30:23] i think. where do I check? [18:30:29] nuria, what do you mean? [18:30:57] madhuvishy: try to log in here: deployment-eventlogging02.eqiad.wmflabs [18:31:17] mforns: did you check the login from mysql consumer is what you expected? [18:31:33] nuria, yes! [18:31:38] mforns: ok, good [18:31:51] mforns: i looked ta it and it seem fine but just to triple chcek [18:32:02] *at it [18:32:14] it works perfectly, the event timestamps are almost identical to the consumer timestamps, all the time [18:32:42] nuria, so it seems a db problem [18:33:00] nuria, I've observed a pattern in the master-slave replication lag [18:33:05] mforns: db not receiving records? [18:33:13] nuria: I think I need to set up something else first. ProxyCommand? [18:33:27] it *seems* to me that the replication lag gets progressively bigger with time [18:33:30] madhuvishy: did you check the ssh config on onboarding page? [18:33:35] nuria, like 10 minutes increase per hour [18:33:40] Yeah seeing that now. [18:33:42] mforns: ahhhhhh [18:33:59] madhuvishy: ok, give it a try [18:34:12] nuria, until the data gap happens, and then the replication lag goes to less than 1 minute [18:34:13] mforns: and - as lag increases- so does the gap? [18:34:21] mforns: what? [18:34:35] mforns: springle is going to like that #not [18:34:45] nuria, the gap in data happens only once or twice per day [18:35:14] nuria, the thing is, I think the gap in data coincides with the replication lag going back to normal [18:35:36] mforns: so funny, in a ahem not-so-good-way [18:36:03] hehe [18:36:27] mforns: sounds like the master is unable to write when replication is being "speed up" or doing more heavy stuff [18:37:34] mforns: ok, then you can work on that with springle, i cannot think of an alarm on our end on that [18:38:10] nuria, yes I agree, the only thing that goes against this is springle's statement, that "the db did not see any inserts of the missing data" [18:38:25] nuria, springle concluded that greping the binary logs [18:39:16] nuria, but yes, I think the best will be reply to him telling about the replication pattern and giving the event examples that he asked for [18:39:59] mforns: ya, also, if you think when next event is going to happen you can "listen" to inserts with tcpdump and see those going to the box i think [18:40:04] ottomata: I am getting diner [18:40:09] Will be back soon [18:40:41] nuria, good idea [18:40:54] mforns: sounds to me that db acknowledges inserts and they are not actually being executed (and thus not being written to db transaction logs) but again i know next to nothing about inners of db setup [18:41:22] nuria, I've read about some innodb insert buffer [18:41:39] nuria, but my understanding of that was so short, that I stopped reading [18:41:49] nuria: something something Permission denied (publickey). [18:42:20] nuria, what? [18:42:26] madhuvishy: ok, with the right ssh config? [18:42:58] mforns: tcdump to listen on port 5000 9incoming/outgoing) [18:43:04] https://www.irccloud.com/pastebin/ojV7NUj0 [18:43:22] mforns: you should be able to listen on mysql port (port is on config) [18:43:40] nuria, ok [18:44:33] madhuvishy: and you did do this right? https://wikitech.wikimedia.org/wiki/Help:Tool_Labs/Access#Creating_a_Labs_account_on_Wikitech [18:44:53] nuria: yup [18:45:05] madhuvishy: BTW, please update the onboarding docs that might be incorrect, the idea is that each hire keeps them current [18:45:08] my public key matches with the one on labs [18:45:21] nuria: okay sure [18:45:36] https://www.irccloud.com/pastebin/KcAlsB4w/.ssh%2Fconfig [18:46:38] are you guys watching 'we're all remoties' talk? [18:46:46] madhuvishy: since this is outside our control, try to ask in #wikimedia-labs and see if you get help there [18:46:55] mforns: arghhh, forgot! [18:47:04] starts in 10 [18:47:08] mforns: do you have the link handy? [18:47:12] mmm [18:47:37] hi madhuvishy nuria [18:47:44] someone needs to add her to the deployment-prep project [18:47:49] (her name is not listed at https://wikitech.wikimedia.org/wiki/Nova_Resource:Deployment-prep) [18:48:01] nuria, no link yet [18:48:05] YuviPanda: someone being ....ahem.... [18:48:23] YuviPanda: a handsome ops indian guy? [18:48:26] nuria: you, for instance :) [18:48:36] or anyone in that list of ‘admins’ :) [18:48:50] YuviPanda: me? wait i do not even know what that is but checking ... [18:48:59] nuria: :D you got admin there :D [18:49:08] nuria: you can go to https://wikitech.wikimedia.org/wiki/Special:NovaProject and add people to the deployment-prep project [18:50:23] nuria: also I don’t think the ops team has any ‘handsome’ indian guys :) [18:50:25] * YuviPanda slinks away [18:50:40] YuviPanda: true that. :P [18:51:05] YuviPanda: finding myself to be admin of things that i barely know what they are is unsettling but let me add madhu there [18:51:56] nuria: :D deployment-prep admin is given out fairly freely [18:52:02] you also have sudo there on all machines for example [18:53:19] madhuvishy: added you to : https://wikitech.wikimedia.org/wiki/Nova_Resource:Deployment-prep, try again [18:55:38] nuria: YuviPanda same error. [18:56:06] YuviPanda: any ideas? [18:56:49] nuria: did you add her at https://wikitech.wikimedia.org/wiki/Special:NovaProject [18:57:03] I don’t see her there [18:58:02] joal, phone call over! [18:58:04] want to batcave? [18:58:37] YuviPanda: but i do not even see where to add people there... [18:58:51] I can help in an hour or two maybe, in the middle of another migration, sorry [18:58:55] ask in -labs? [18:59:32] nuria: actually, let me just add her [18:59:41] YuviPanda: sounds good, ok [18:59:47] madhuvishy: nuria done [18:59:55] but you should update your onboarding docs :) [19:00:08] YuviPanda: thank you, madhuvishy please try again, if it no work we shall ask in labs [19:00:41] YuviPanda: had just done it: https://wikitech.wikimedia.org/wiki/EventLogging/Testing/BetaLabs#Give_people_access [19:00:48] cool :) [19:01:25] YuviPanda: i am all about docs otherwise i forget EVERYTHING [19:02:04] nuria: YuviPanda thanks, it seems to have worked. milimetric pointed out I had <> around my username in config. [19:02:34] ottomata, so: how come HDFS is so backlogged? [19:02:41] Like, I get the production tasks, but this is..new. [19:02:49] madhuvishy: all right, one less thing, i want to watch the "all remoties" talk (that i think has not started yet) [19:02:59] madhuvishy: we can go to batcave after [19:03:05] nuria: me too :D [19:03:16] nuria: sure. i will also get lunch. [19:03:20] ottomata: joining the remoties neeting [19:03:27] madhuvishy: make sure to install navigationtiming extension and that your mw instance is running on http://localhost:8080 [19:03:28] Can chat thoughb [19:03:39] nuria: done [19:03:51] madhuvishy: open your network panel [19:04:07] madhuvishy: everytime you load that page you should see a req trying to hit bits [19:04:26] haha, ok, so , i'm not sure this is 100% fixed now...but [19:04:26] that comes back with an error (as bits is not setup on vagrant0 [19:04:35] remember that job I mentioend yesterday that was taking forever? [19:04:35] ottomata: yesssss...... [19:04:49] it was messed up because of analytics1020 [19:04:52] nuria: yes that seems to be happening [19:04:56] yesterday, cmjohnson fixed an20 [19:05:06] but, wasn't quite done fixing it, and had left hte network cable unplugged for something [19:05:17] so, some containers got allocated there [19:05:18] madhuvishy: ok, that is the "client" of eventlogging, javascript code that lives on mediawiki [19:05:23] but an20 disappeared [19:05:27] i dunno why yarn didn't fix itself [19:05:32] but those containers were just stuck in timeout [19:05:35] waiting for response [19:05:42] weird [19:05:44] the app master just sat there doing nothing [19:05:44] madhuvishy: we own both client and server (and both are on same depot on EventLogging extension) [19:05:57] which took up a lot of the fair share, because ti still had resources allocated to it [19:05:57] nuria: right, okay.. [19:06:09] anyway, i killed that, and am trying to get prod jobs caught up [19:06:18] webrequest load jobs are doing well [19:06:18] still alot in the essential queue ottomata [19:06:26] once i saw load jobs were cool [19:06:26] yeah [19:06:33] ok [19:06:36] i unsuspended refine jobs [19:06:41] now things are looking pretty slow again [19:06:44] the one thing that is werid [19:06:45] madhuvishy: in vagrant the python server code is not deployed [19:06:53] is that the first job in the queue takes up most of the fair share [19:06:56] madhuvishy: but you can run unit tests and execute some of it [19:06:57] and things still don't get allocated [19:07:06] joal, our essential queue is fifo [19:07:10] i'm thinking about changing it to fair. [19:07:22] hmm [19:07:23] madhuvishy: see that you can run tests: http://www.mediawiki.org/wiki/Extension:EventLogging#How_to_run_tests [19:07:33] and also making another queue [19:07:34] . [19:07:37] or, maybe we keep eseential fifo [19:07:41] and make a production queue [19:07:42] that is fair [19:07:45] Wouldn't it be better to keep it fifo, and add other queues ? [19:07:48] yeah [19:07:49] maybe so [19:07:50] yup [19:07:57] so queues. [19:08:23] default [19:08:23] _something else in between default and production_? [19:08:23] production [19:08:23] essential [19:08:24] ? [19:09:28] madhuvishy: will be watching talk, let me know if you run into trouble running tests [19:09:52] madhuvishy: it might be that php unit tests docs needs updating [19:10:32] default < production < essential < top_prio (to catch up on errors for instance) [19:10:38] ottomata: --^ [19:11:21] hm, naw, cause ellery and leila wanted a queue that they could submit things to that was > default, but not production [19:11:34] and, if we move more jobs out of essential [19:11:38] essential will be top prio [19:11:40] Yeah, but how to name that ? [19:11:46] i'm thinking essential just has [19:11:47] camus [19:11:49] load [19:11:52] mayyybe refine [19:11:55] ? [19:12:13] I think refine can go in production [19:12:17] aye ok. me too [19:12:22] Doesn't need to prehempt [19:12:37] aye, i mean, all the other jobs will depend on it. [19:13:02] Yes, but we don't loose data if it doesn't run [19:13:30] But it's one of the important, correct [19:13:50] yes, i mean, we don't lose data if load doesn't run [19:13:51] only if camus doesn't run [19:13:57] ottomata: in that thread, we were asking for access to essential for a period of time, but if doing it in a different and more permanent way makes more sense, we're definitely open to that. [19:13:57] nuria: alright. running them. brb lunch [19:14:09] leila, it does make sense [19:14:12] help us think of a name [19:14:15] it is nto production [19:14:18] itis not default [19:14:20] what is the name of the queue? [19:14:23] privileged? [19:14:24] naw [19:14:55] humm. thinking. [19:14:55] ottomata, you mean root.essential? [19:14:55] bonus! [19:14:56] ? [19:14:58] ottomata: so it's either camus + load + refine in essentials, or camus only [19:15:04] or, to distinguish from root.essential? [19:15:07] ottomata: Jeopardy ! [19:15:08] new queue [19:15:11] Ironholds: [19:15:23] aha [19:15:34] joal: load is important toooooo, dunno, i could see camus and load in essential [19:15:46] so if essential is for the root, shit will explode and everything falls over if this doesn't work [19:16:04] ottomata: yeah, but as you said as well, refine is core to users [19:16:11] root.prioritised? [19:16:12] nothing happens with load only [19:16:12] ottomata, joal, so what you have in mind will go between default and production in terms of priority? [19:16:21] prioritised! is prettty good [19:16:24] correct leila [19:16:24] default, prioritised, essential [19:16:31] ottomata: I like it [19:16:35] + production [19:16:42] produciton > prioritized [19:16:46] does that make sense? [19:16:54] default < prioritized < production < essentials [19:16:57] a [19:16:58] ja [19:17:03] +1 to prioritized as long as it has a clear distinction from top_prio [19:17:05] priority* [19:17:58] ottomata: so camus + load + refine in essentials, FIFO, with maybe less prehemption than currently ? [19:18:05] okay, so we're dropping top_prio in [19:18:05] default < production < essential < top_prio [19:18:05] and replacing it with [19:18:05] default < prioritized < production < essentials joal, ottomata? [19:18:14] goot to me :) [19:18:15] yup [19:18:32] hmm, ja what should preemption be? i dunno. maybe we shoudl jsut put camus in essential. it shoudl preempt! [19:18:36] but maybe the other two shoudlnlt [19:18:54] Yeah, that was my point [19:18:54] the logic works well for me, as long as in practice the prioritized works well for us. [19:19:04] leila: ;) [19:19:13]