IRC logs for #baserock for Thursday, 2017-06-22

*** radiofree is now known as RadioFree00:39
*** RadioFree is now known as RADIOfree02:12
*** RADIOfree is now known as radiofree02:17
*** gtristan has joined #baserock05:44
*** noisecell has joined #baserock08:09
*** gtristan has quit IRC08:10
*** gtristan has joined #baserock08:15
*** edcragg_ has quit IRC08:33
*** edcragg has joined #baserock08:36
*** ssam2 has joined #baserock08:44
*** ChanServ sets mode: +v ssam208:44
paulsherwood/win/win 808:45
paulsherwoodbuh08:45
*** jonathanmaw has joined #baserock08:50
*** bruce has joined #baserock09:02
*** bruce is now known as Guest5151909:03
*** jude_ has joined #baserock09:38
*** locallycompact has joined #baserock09:52
*** CTtpollard has quit IRC09:59
*** CTtpollard has joined #baserock11:10
ssam2jjardon, do the Baserock gitlab runners have local cache that persists between runs?11:22
ssam2or does it clone gits for every build ?11:22
*** gtristan has quit IRC11:25
*** tiagogomes has quit IRC11:32
*** gtristan has joined #baserock12:05
*** tiagogomes has joined #baserock12:22
jjardonssam2: no cache between runs: they are ephemeral so to persist the cache  Distributed runners caching needs to be configured and enabled: https://gitlab.com/gitlab-org/gitlab-ci-multi-runner/blob/master/docs/configuration/autoscale.md12:41
ssam2right so we'd  need an S3 bucket ? :-(12:48
ssam2would be nice if https://docs.gitlab.com/ce/ci/yaml/README.html#cache would work12:49
jjardonssam2: yes, that or have runners always up12:52
ironfootwe could have something always up in Datacentred12:54
ironfoot"always" :P12:54
ironfootbut if this "somethin" needs to be a runner, I guess that the builds won't be as fast as they are right now12:55
ssam2well, if we get BuildStream builds that are using the Freedesktop SDK host tools and have no cache, they will take hours12:58
ssam2as in my experience it takes hours to pull the thing12:58
ssam2maybe if they're running in AWS it'll be faster though12:58
jjardonironfoot: happy to setup a runner in Datacentred if a machine is provided12:59
ssam2plus we'd be cloning linux.git and gcc.git each time which is slow as feck12:59
ironfootIt will be easy for me to allocate some resources in Datacentred, not in AWS though13:00
jjardonssam2: Can Buildstream build everything from scratch, instead using the Freedesktop SDK?13:21
ssam2no, it has to have some prebuilt host tools13:23
ssam2i've been using Alpine Linux locally which is just over 200MB for a toolchain etc.13:24
ssam2i'm not sure how big Freedesktop SDK is, but it sure takes a while to pull from OSTree for me13:24
ssam2Baserock should probably one day produce its own "host tools"13:25
ssam2which would basically be the stage2-*.morph elements13:25
ssam2so effectively we'd bootstrap the same way, it's just arranged differently13:25
ssam2we'd need CI to build tagged versions of the stage2.bst stack and push it to a public OSTree repo13:26
ssam2then Baserock definitions would refer to that OSTree repo as the first dependency of each build13:26
ssam2so it's like what we did before with stage2-* chunks in a kbas / morph-cache-server cache... except without the mess of bootstrapping on a variety of hosts13:27
ssam2which goes badly wrong when the GLIBC ABI changes13:27
jjardonI see13:41
gtristanartifact cache is coming soon13:43
gtristanAnd what we're going to want, which I'm not sure is going to be an issue or not...13:44
gtristanIs an ssh key pair13:44
gtristanjjardon, maybe you know about this for the runners, but... what we want is to collect the public keys of anyone who can push artifacts to the server13:44
gtristanjjardon, that would mean the runner needs a private key, and that private key should remain private (i.e. it should not be revisioned in git and copied into place in a .gitlab-ci.yml, it should be a part of the host that is auto-building)13:45
gtristanI'm not sure how best to set that up, or if it's possible on gitlab runners13:45
*** gtristan has quit IRC13:59
jjardongtristan: maybe generate and use a docker image with those keys included?14:06
jjardonbut that is very inflexible, maybe using deploy keys? a possible solution could be this: https://docs.gitlab.com/ee/ci/ssh_keys/README.html14:09
ssam2that looks like what we would need14:10
ssam2GitLab EE only though ?14:10
benbrown_not EE only14:13
ssam2ah, sound then14:14
benbrown_Potentially stupid question, but is population of the bst remote cache just through scp?14:16
* benbrown_ is wondering how much sanity checking there is likely to be on the receiving end14:16
ssam2through sshfs I believe14:19
*** gtristan has joined #baserock14:21
*** toscalix has joined #baserock15:27
*** Guest51519 has quit IRC15:38
*** Guest51519 has joined #baserock15:38
*** jonathanmaw has quit IRC16:36
*** jude_ has quit IRC16:55
*** noisecell has quit IRC17:11
*** ssam2 has quit IRC17:16
*** toscalix has quit IRC17:25
*** locallycompact has quit IRC18:18
*** tiagogomes has quit IRC18:35
*** paulwaters_ has joined #baserock19:31
*** locallycompact has joined #baserock20:38
*** locallycompact has quit IRC20:38
*** Guest51519 has quit IRC22:41

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!