IRC logs for #baserock for Friday, 2016-02-19

*** edcragg has quit IRC01:06
*** gtristan has quit IRC05:29
*** gtristan has joined #baserock05:40
*** toscalix has joined #baserock07:40
*** CTtpollard has joined #baserock08:01
*** edcragg has joined #baserock08:12
paulsherwood16-02-19 06:57:51 [372/447/447] [WebKitGtk] Elapsed time for build of WebKitGtk.eb8b1b427036d5bfa18f1e82b554f3310815a238638ad122c170219cad89b03c 05:40:3408:47
paulsherwoodneed moar cpu :)08:47
*** edcragg has quit IRC09:09
*** tiagogomes_ has joined #baserock09:13
pedroalvarezpaulsherwood: ow, how fast is the machine you used to get that elapsed time?09:15
pedroalvarezjust asking because a 2vcpu VM is getting the same elapsed time (i believe, I'll double check)09:16
rjekWebKit doesn't so much need more CPU, but less code :)09:16
pedroalvarez2016-02-06 02:44:57 Started building WebKitGtk on mason-x86-64:343409:17
pedroalvarez2016-02-06 07:35:56 Progress: Transferring WebKitGtk-misc to shared artifact cache09:17
pedroalvarez2016-02-06 07:37:12 Finished building WebKitGtk on mason-x86-64:343409:17
paulsherwoodmwilliams_ct: ^^ ?09:20
mwilliams_ctI assume this is running on c.b.o?09:21
* mwilliams_ct will check specs09:21
pedroalvarezchiken bacon and onion09:21
* pedroalvarez goes for a breakfast09:21
pedroalvarezand for a missing "c" in chicken09:22
gtristanpedroalvarez, I hope it's organic :D09:22
* gtristan goes for supper09:22
pedroalvarezgtristan: always!09:23
mwilliams_ctthe workers for Concourse are m3.large EC2 instances, has the specs. 2vcpu, 7.5GiB ram, 32GB ssd storage09:23
mwilliams_ctwe could easily spin up bigger machines if needed, it's just a quick tweak to the bosh manifest (though obviously that incurs a higher cost so we should be sure it will help)09:24
pedroalvarezand I take that the workers have to use docker to run the instructions of the pipeline, correct?09:25
*** ssam2 has joined #baserock09:26
*** ChanServ sets mode: +v ssam209:26
mwilliams_ctpedroalvarez: yes, everything runs in containers (I feel so hipster typing that)09:27
pedroalvarezcall them "cans" and you will be allright09:27
*** gtristan has quit IRC09:28
*** edcragg has joined #baserock09:30
pedroalvarezBtw, yesterday I found an interesting bug in git-review..09:34
pedroalvarezI reported it upstream, but I doesn't look like they really use the bug-tracker09:34
pedroalvarezIf you try to send for review a branch that is already in gerrit repo, it will say that "submitting the change would be pointless"09:37
ssam2ah, yeah09:37
ssam2i'm not sure if it's quite a bug09:38
ssam2definitely annoying & confusing though09:38
pedroalvarezwho will push branches to gerrit? nobody, but we push to the trove, that gets mirrored into gerrit09:38
edcraggsubmission is futile09:39
ssam2maybe if you pass --track or give the name of a target branch then it works correctly?09:40
ssam2if not, definitely seems like a bug09:40
pedroalvarezoh, thanks09:43
pedroalvarezwill try to use that09:43
paulsherwoodmwilliams_ct: it would definitely help. but will they spin DOWN again when not in use?09:43
* paulsherwood gets irritated when we leave all the lights on for no good reason09:44
mwilliams_ctpaulsherwood: no09:44
paulsherwoodwell, that's a bug :)09:44
mwilliams_ctthey don't automatically spin down, and I still havent found a way to do that in bosh09:44
paulsherwood=> concourse, bosh, the world? :)09:44
pedroalvarezno results for concourse+elastic+agents :(09:47
mwilliams_ct might work, I could look into it09:48
*** jonathanmaw has joined #baserock09:48
tiagogomes_Is everyone happy but me that the submodules on the emailing list is not generic enough to handle projects with multiple sources that aren't submodules?09:52
* paulsherwood can't believe we're the only people in the world that don't want to throw money away :)09:52
paulsherwoodtiagogomes_: i don't understand, sorry09:52
ssam2tiagogomes_: I'm happy that we're making some progress, and I don't think Dan's proposal blocks any future proposal to make things more general09:53
paulsherwoodif it works for all of the things we currently have in definitions, i'd say it's worth having.09:53
pedroalvareztiagogomes_: I think I'm with you, but din't have time to think deep about it09:53
pedroalvarez(or deeply)09:53
paulsherwoodwe've suffered previously from attempting to solve all cases when real-world requirements are specific09:54
pedroalvarezi remember doing some "hacks" in some chunks, like commiting all the sources of another repo, to make it work09:55
jjardonssam2: +109:55
tiagogomes_ssam2 right, but I'd favor a final solution than an intermediate solution which will require to be reworked in the future. If we change something and later on rechange the same thing, it gives me the idea that we haven't got a plan and we don't know where we are heading09:56
paulsherwoodtiagogomes_: i think you're wrong in that09:57
paulsherwoodas follows:09:57
paulsherwood1) perfect is the enemy of the good09:57
paulsherwood2) small steps roughly in the right direction are better than no steps while we think a lot about big steps09:57
paulsherwood3) definitions are yaml - we can automate migration for any future work09:58
paulsherwood4) "No plan survives contact with the enemy."09:58
paulsherwoodoh, and 'final solution' is an overloaded phrase, best not used in polite company :)09:59
*** rdale has joined #baserock10:02
tiagogomes_1) I didn't mention anything about perfection, but we should handle the case of multi-sources which aren't submodules. 2) Is a step that you will need to redo at some point 3) It will still introduce extra cognitive work in the user to understand definitions 4) No comments.10:02
paulsherwood1) do we actually have any?10:02
paulsherwood2) we typically end up redoing things anyway :)10:02
paulsherwood3) agreed10:02
paulsherwoodtiagogomes_: are there any current/known cases of 'multi-sources which aren't submodules' ?10:04
paulsherwood(in definitions)10:04
pedroalvarezfor 1) I found one case. we copy in rabbitmq-codegen.morph the sources to DESTDIR to use them later in rabbitmq-server10:05
pedroalvarezfor gcc we commited the sources to the repo directly I believe10:05
tiagogomes_paulsherwood, GCC for example, which depends in some math libraries:
paulsherwooddo we have any actual fix in mind for this?10:06
paulsherwoodmaybe locallycompact could be persuaded to improve/extend his soln10:06
pedroalvareztiagogomes_: what I don't understand is why the submodules solution won't handle this cases too10:07
pedroalvarezI mean, the migration script won't, but that doesn't stop us from using the new feature to put more sources in the staging area10:08
richard_1awAIUI submodules would work, but that requires more local changes to get things working, when the usual workflow for gcc is bring these sets of components together externally10:08
*** richard_1aw is now known as richard_maw10:09
*** gtristan has joined #baserock10:12
*** tiagogomes_ has quit IRC10:13
edcraggi had a bsp which required a set of scripts from one repo to be used on the source from another repo, for that i copied stuff to /tmp in one chunk, and copied it back in the build for the later chunk. probably an isolated example, seemed like a bit of a hack at the time10:20
*** tiagogomes_ has joined #baserock10:22
pedroalvarez"a bit of a hack" :)10:22
*** Lachlan1975 has joined #baserock10:25
tiagogomes_pedroalvarez, I haven't looked at the patches closely. But then it shouldn't be called submodules. Also is it possible to have multiple top levels?10:34
*** locallycompact has joined #baserock10:34
paulsherwoodlocallycompact: ^^ ?10:35
locallycompactpaulsherwood, ^^ to you too.10:35
* locallycompact assumes there's context10:35
* edcragg can't see tiagogomes_'s last comment at the end of that log10:40
*** Lachlan1975 has quit IRC10:40
edcraggseems to be there now10:40
locallycompactto be clear, this is *only* submodules, you can't use this for the gcc thing. It's translating submodules it finds in .gitmodules using the key values in definitions.10:42
paulsherwoodedcragg: logging takes time ;)10:43
pedroalvarezthat's the migration script, of course10:43
locallycompactpedroalvarez, no, that's the behaviour of ybd10:43
pedroalvarezoh, ok10:43
locallycompactif you add submodules to a project that doesn't have any it doesn't do anything10:44
pedroalvarezoh well10:44
pedroalvarezanyway, I'm not going to block this10:44
pedroalvarezlooks useful to me10:44
locallycompactthe gcc thing is quite out of my way. This change is so I can lorry things that have submodules and not have to worry about branching every time10:44
paulsherwoodand +1 for this change, from me :)10:45
ssam2I would prefer +1s and -1s on the mailing list10:46
locallycompacthow do I clone the baserock wiki10:47
locallycompactnvm got it10:48
paulsherwoodssam2: done10:52
edcraggtiagogomes_: by multiple top levels, did you mean multiple clones per chunk by any chance?10:55
*** Lachlan1975 has joined #baserock10:55
ssam2paulsherwood: thanks. I take it you implicitly support the proposal itself?10:56
ssam2I would like us to consider the actual proposal as well as the code. Since the definitions format should exist independently of any tool's codebase10:56
paulsherwoodyes, i do. i'm worried if this will break for morph users, though10:56
ssam2it will if we migrate master of the reference systems to definitions version 8 before Morph can understand version 810:57
tiagogomes_edcragg yes, something like this
edcraggtiagogomes_: exactly what i was thinking, i think something like that could be useful10:58
locallycompactIf we migrate any actual versions of things that have submodules then it will break for morph, those changes by itself won't cause morph to fail iiuc10:59
ssam2what do you mean by "those changes" ?11:00
locallycompactmy series11:00
ssam2patch 2/2 would break Morph11:01
locallycompacthow so?11:02
paulsherwoodhave we tried it?11:02
ssam2ok, I haven't tried it11:02
ssam2I will do11:02
ssam2but it will say that it doesn't support definitions version11:02
locallycompactOh right11:02
ssam2ideally we'd hold off on that part, and someone will soon add support for Morph to definitions v811:03
ssam2if nobody does within a reasonable time then, I don't know, maybe Morph is dead11:04
ssam2that would be a shame though, not just because it has useful features YBD lacks, but also because having 2 implementations of the definitions format catches a lot of problems11:07
*** dabukalam has joined #baserock11:12
locallycompactI think I successfully pushed to gerrit11:38
locallycompactJust the first part reformatted + the schema11:38
paulsherwoodlocallycompact: you've included fix to json-schema stratum was that intentional?11:40
locallycompactI believe that was what ssam2 asked for on list11:40
paulsherwoodlet's see if he's +1 too :)11:44
jjardonpaulsherwood: sorry, forgot to tell, the fix is available here:
jjardonrichard_maw: ok to remove your -1 in ?12:20 will be down for a few minutes12:38
ssam2ok, it's back and with added syntax highlighting :-)12:40
locallycompactit's doing a strange with the page links12:41
ssam2what's wrong?12:42
locallycompactyou don't see that?12:42
locallycompactsec i"ll screenshot it12:42
ssam2oh, I see that12:43
ssam2maybe a cgit bug12:44
richard_mawjjardon: better than that, I'll +212:52
jjardonrichard_maw: thanks! :)12:53
ssam2weird... /var/www/htdocs/cgit/cgit.css on doesn't match the CSS file it's actually serving12:54
ssam2hence the <li> elements are lacking the style that makes them all appear on the same line12:54
ssam2restarting lighttpd-git.service has fixed it12:55
ssam2I guess the lighttpd-git systemd service is missing a dependency on local-fs.target12:56
pedroalvarezI was looking at the trove I did the testing on12:57
pedroalvarezit was ok in there12:57
*** gtristan has quit IRC13:27
*** gtristan has joined #baserock13:36
jjardonssam2: pedroalvarez we are missing commits: gerrit said they are merged but it doesnt appear in git.baserock.org14:33
jjardon(python3-core patch for example)14:33
ssam2ah, ok14:58
ssam2thanks for spotting it14:58
ssam2seems the commits are in gerrit's master branch but weren't replicated to yet15:01
jjardonalso, I m not sure im getting email notifications from gerrit anymore?15:03
ssam2i am15:04
paulsherwoodjjardon: i'm trusting you on :)15:08
ssam2gerrit-replication seems totally wedged, I'm going to try restarting gerrit, which takes a few minutes15:11
jjardonssam2: too many patches? :)15:12
jjardonssam2: or maybe we should stop mirroring to g.b.o  and activate cgit in gerrit?15:16
ssam2seems the ssh host keys for have changed, which is a bit broken since I did an upgrade15:16
ssam2well, not too difficult to change them back15:17
pedroalvarezlooks weird that ybd honours upgrade-type but not upgrade-location15:21
paulsherwoodso in theory it should15:25
paulsherwoodi wonder what parent_location is15:26
paulsherwoodSotK: ???15:26
richard_mawpaulsherwood: assuming it's got the same meaning as in morph, that's the location of the deployment it's a subdeployment of, so you can have a deployment deploy a file inside another deployment15:27
richard_mawe.g. making an initramfs15:27
SotKrichard_maw is correct as far as I remember15:30
gary_perkinsHi! I have a ybd issue: I'm trying to perform an update to a trove. The update complained about no VERSION_LABEL, so I supplied "trove.VERSION_LABEL=2016-02-19" on the cmd line. ybd didn't like that, so I added VERSION_LABEL: 2016-02-19 to the cluster file, then... :(15:31
gary_perkinsHas anyone had that before?15:31
paulsherwoodgary_perkins: put quotes around it, please15:32
paulsherwoodlooks like python is converting that to a date15:33
gary_perkinspaulsherwood: ahh!15:33
paulsherwoodor use a jrandom string. it's a bug, though15:33
pedroalvarezpaulsherwood: regarding upgrade-location, the line triggering the previous error is
gary_perkinspaulsherwood: thanks, it appears to be proceeding now :)15:34
paulsherwoodpedroalvarez: yup, hence i was desperately clutching at SotK's ankles15:36
* paulsherwood has never fully understood deployment15:36
paulsherwoodpedroalvarez: could you put in 'import pdb; pdb.set_trace()' at line 69 of and see what things are being set to?15:39
paulsherwoodnm, i'll have a play over the weekend15:46
*** CTtpollard has quit IRC16:03
pedroalvarezyeah better, because it wasn't happening to me  :/16:11
*** brlogger has joined #baserock16:25
ssam2testing ircslogs16:25
gary_perkinsok, the updated trove is booting into single-user-mode without nary an explaination as to why. The only error message I can see is about /etc/init.d/rcS not existing. But other troves don't have that file either!16:34
gary_perkinsthe journal has no entries either :(16:35
gary_perkinsanyone had this before?16:35
pedroalvarezI have never seen that error16:36
gary_perkinslooks like I'll have to revert back to the older configuration for now16:37
pedroalvarezI hope you can do that16:37
ssam2fwiw, i've deployed two working troves from master of the reference system definitions.git16:37
ssam2well, from infrastructure.git having merged master of definitions.git16:37
* gary_perkins selects "factory"16:37
gary_perkinsit boots16:38
ssam2i might be able to have a look in a bit, just need to get backups working again16:38
radiofreehasn't it been upgraded since initial deployment?16:38
gary_perkinsssam2: yeah, no worries, ta16:38
pedroalvarezradiofree: nope, first upgrade AFAIK16:38
* gary_perkins nods16:38
ssam2s/deployed/upgraded/ in my comment16:39
radiofree/etc/init.d/rcS doesn't sound very systemdey16:41
radiofreealso isn't there some issue in baserock where the journal doesn't give you anything?16:44
radiofreesomething to do with /var/run... or /run/var...16:44
ssam2oh yeah, journal can be broken because /var is a subvolume16:45
radiofreethat probably should be fixed...16:46
richard_mawit's fixable by making / (and hence /var) read-only at boot-time, so systemd spools its logs in /run until /var has been mounted properly16:48
richard_maw(note, Lennart wants to replace this logic with requiring the initramfs mount /var, but he's getting no traction)16:48
radiofreewhat's the default "storage" behaviour?16:49
radiofreeshouldn't "persistent" fallback to /run/?16:49
radiofreeah sorry, that, misread your comment16:51
ssam2all infra should be up to date now, if you see anything wrong, please let me know17:00
*** franred has quit IRC17:43
*** ssam2 has quit IRC17:43
*** jonathanmaw has quit IRC17:43
*** tiagogomes_ has quit IRC17:54
*** rdale has quit IRC17:54
* pedroalvarez continues investigaing "raise RuntimeError("generator didn't stop")" error in --quiet mode18:03
pedroalvarezso.. if a context manager has the yield in a loop, it becomes a generator.. right?18:07
pedroalvarezoh well... I don't understand why it fails in openstack.write and it doesn't in rawdisk.write18:12
richard_mawno, it becomes a faulty context manager18:12
pedroalvarezthat's good to know18:12
pedroalvarezthen maybe find_and_mount is one of those18:12
richard_mawcontext managers *must* only yield *once*18:12
*** edcragg has quit IRC18:13
pedroalvarezthanks for the hint richard_maw18:17
pedroalvarezfound the problem and "the solution"18:17
pedroalvarezI'll look into a nicer one when I have some time18:17
*** bfletcher has quit IRC18:25
*** bjdooks has quit IRC18:25
*** cyndis has quit IRC18:25
*** JPohlmann has quit IRC18:25
*** mwilliams_ct has quit IRC18:25
*** JPohlmann has joined #baserock18:25
*** JPohlmann has joined #baserock18:25
*** bjdooks has joined #baserock18:25
*** mwilliams_ct has joined #baserock18:25
*** cyndis has joined #baserock18:25
*** rjek has quit IRC18:27
*** rjek has joined #baserock18:28
*** toscalix has quit IRC18:28
*** rjek has quit IRC18:31
*** rjek has joined #baserock18:31
*** locallycompact has quit IRC18:31
*** Lachlan1975 has quit IRC19:20
*** bfletcher has joined #baserock19:28
*** locallycompact has joined #baserock20:00
*** locallycompact has quit IRC21:57

Generated by 2.14.0 by Marius Gedminas - find it at!