ChanServ changed the topic of #freedesktop to:
ngcortes has quit [Remote host closed the connection]
blue__penquin has joined #freedesktop
imirkin has joined #freedesktop
Kayden has joined #freedesktop
veganaiZe has quit [Quit: veganaiZe]
ximion has quit []
blue__penquin has quit [Quit: Connection closed for inactivity]
alanc has quit [Remote host closed the connection]
alanc has joined #freedesktop
i-garrison has quit []
i-garrison has joined #freedesktop
i-garrison has quit []
i-garrison has joined #freedesktop
i-garrison has quit []
i-garrison has joined #freedesktop
bcarvalho has quit [Quit: Leaving]
<bentiss> notice: a security update of gitlab showed up, I am going to apply it this morning
jpnurmi has quit []
jpnurmi has joined #freedesktop
<bentiss> and here we start, expect 502 and some 404 when accessing git data while the various pods get updated and restarted
<bentiss> 502s should be over now
laguneucl has joined #freedesktop
<i-garrison> bentiss: I get a 500 hope that's ok
<bentiss> one node is not happy :(
<bentiss> i-garrison: ^^
<bentiss> it's back (the node)
bcarvalho has joined #freedesktop
<i-garrison> bentiss: all good now, thanks !
<bentiss> all the services are up now according to k8s
<bentiss> and the dissks are happy again
<bentiss> i-garrison: well, I just waited for kubernetes to sort his problem out by itself :)
<i-garrison> bentiss: .. except for this glitch, https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/598 has a duplicate for commit I happen to push during change
<i-garrison> bentiss: git is ok, I think it is just the discussion page glitch
<bentiss> i-garrison: if it's just in the discussion, that should be fine
<bentiss> probably the db was lagging behind and did not set a lock
<bentiss> (AFAICT, the db is on the node that was having troubles)
<bentiss> doh!, last month GCP bill is just over $1000 -> $1076.71
<bentiss> daniels: ^^ one thing I noticed is that the way the runners are set, they are still polling the image from GCS even if they have an override for the registry cache
<bentiss> daniels: maybe we can tell ci-templates (or projects) to do a s/registry.fd.o/registry-mirror.fd.o/ in the 'image:' entry of the jobs
<bentiss> this way the images will only be pulled by k3s and served from there
danvet has joined #freedesktop
<MrCooper> not possible to do that transparently behind the scenes?
blue__penquin has joined #freedesktop
<MrCooper> bentiss: looks like pushing to the mirror fails: https://gitlab.freedesktop.org/daenzer/mesa/-/jobs/11470369
<bentiss> MrCooper: the mirror is read-only, it doesn't have the credentials to push
<MrCooper> k, makes it a bit more complicated
<bentiss> MrCooper: I tried doing it transparently by overriding the IP on the runners, but the gitlab runner helper is running in a docker container and checks for the image and pulls it from the original server
<bentiss> MrCooper: my idea is to replace in ci-template (or in the projects for those that overwrite it) the image field of the job
<bentiss> but within the job itself, pushes should go to registry.fd.o
<MrCooper> right
<MrCooper> the registry path can't contain any upper case characters, right?
<bentiss> I don't think
<bentiss> MrCooper: on a different topic, I brought up the dependencies for the mesa pipeline (https://gitlab.freedesktop.org/daenzer/mesa/-/pipelines/352002 for example), and I was wondering: why all the panfrost jobs depend on x86_build? Are they cross-compiling?
<MrCooper> the CI jobs themselves run on x86 GitLab CI runners; the ARM devices are separate, controlled via LAVA
<MrCooper> bentiss: so Mesa would need to either do s/registry.fd.o/registry-mirror.fd.o/ on $CI_REGISTRY(_IMAGE), or convert $CI_PROJECT_PATH to lower case; not sure how either could be done for image:
<bentiss> MrCooper: re-panfrost: OK, thanks!
<bentiss> MrCooper: re-mirror: not sure what is the problem here
<bentiss> right... I think I understand the problem
<MrCooper> $CI_PROJECT_PATH may contain upper case characters; $CI_REGISTRY(_IMAGE) doesn't, but contains the wrong domain
<bentiss> yeah, problem is the mesa namespace is 'Mesa' :(
<MrCooper> didn't think so, but both namespace and project name can be anything for forked projects
<bentiss> sigh, I wish we could use https://docs.gitlab.com/ee/ci/variables/#pass-an-environment-variable-to-another-job -> only inside 'script'
<MrCooper> $CI_PROJECT_PATH_SLUG doesn't work either, it contains a dash instead of the slash
* bentiss is wondering if there would be something easier than using dynamic child pipelines
<bentiss> cause that would work, but at a cost...
<bentiss> MrCooper: for starters, the FDO_BASE_IMAGE can easily havve a replace in registry.fd.o by using a before_script
<bentiss> which means that rebuilding images would fetch from the cache
<bentiss> however, not sure how we can do the same for image:
<MrCooper> right, that's pretty rare though I'm afraid
<MrCooper> compared to build/test jobs
<bentiss> FWIW, if we were to do that at the server level, the snippet would have to check which kind of request it is (PUT, PATH, GET or POST) and return a 302 accordingly, which is kind of error prone
<bentiss> and Ideally, we would host our data on packet, but we don't have 7TB (times 3) available
<bentiss> well, we kind of have actually :)
<MrCooper> FWIW, I suspect a big chunk of the docker images are stale and could be removed, but figuring out which could be tricky
<bentiss> yeah, that's why I am waiting on the new db registry backend, but gitlab scheduled a production ready end of this summer I think
<bentiss> cause right now, every time you do a registry query, the server opens a bunch of files on google cloud storage
<bentiss> it is really not optimized for big registries
blue__penquin has quit [Quit: Connection closed for inactivity]
blue__penquin has joined #freedesktop
ximion has joined #freedesktop
reillybrogan_ has joined #freedesktop
reillybrogan has quit [Ping timeout: 480 seconds]
Haaninjo has quit [Quit: Ex-Chat]
asdflkj_sh has joined #freedesktop
asdflkj has quit [Ping timeout: 480 seconds]
ximion has quit []
laguneucl has quit [Ping timeout: 480 seconds]
chomwitt has joined #freedesktop
NoGuest17 has quit [Remote host closed the connection]
veganaiZe has joined #freedesktop
tobtoht has quit [Remote host closed the connection]
blue__penquin has quit [Quit: Connection closed for inactivity]
tobtoht has joined #freedesktop
jarthur has joined #freedesktop
jarthur has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
jarthur has joined #freedesktop
ximion has joined #freedesktop
jarthur has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Haaninjo has joined #freedesktop
jarthur has joined #freedesktop
K`den has joined #freedesktop
Kayden has quit [Remote host closed the connection]
K`den is now known as Kayden
jarthur has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
bcarvalho has quit [Remote host closed the connection]
bcarvalho has joined #freedesktop
austriancoder has quit [Remote host closed the connection]
SanchayanMaity has quit [Remote host closed the connection]
SanchayanMaity has joined #freedesktop
austriancoder has joined #freedesktop
bcarvalho_ has joined #freedesktop
bcarvalho has quit [Read error: Connection reset by peer]