ChanServ changed the topic of #freedesktop to: https://www.freedesktop.org infrastructure and online services || for questions about freedesktop.org projects, please see each project's contact || for discussions about specifications, please use https://gitlab.freedesktop.org/xdg or xdg@lists.freedesktop.org
<pendingchaos>
this channel is mostly for freedesktop infrastructure and online services, rather than hosted projects
<pendingchaos>
I don't know what would be a good channel for d-bus
<daniels>
Harzilein: yeah, the dbus@ mailing list is the best forum for it
ybogdano has quit [Ping timeout: 480 seconds]
damian has quit [Read error: Connection reset by peer]
<bentiss>
MrCooper:all the runners are now using podman
<MrCooper>
meson tests are timing out for no apparent reason
<bentiss>
MrCooper: the "glamor" part makes me think it needs to have access to the host
<bentiss>
and that's bad anyway
<bentiss>
so I would ask to use boot2container to start the tests in a VM
<MrCooper>
not sure what you mean
<MrCooper>
the glamor tests use llvmpipe
<MrCooper>
I saw similar issues yesterday (in Mesa IIRC) with the -11 runner before you took it out
<bentiss>
MrCooper: previously, we had full access to the host with docker as the containers are privileged. IOn the short future I want to switch to rootless podman, so we better fix our CI instead.
<bentiss>
This way we could securely do releases through CI
<MrCooper>
I get that, but I don't see how this issue is related
<MrCooper>
it doesn't have anything to do with docker or host access
<bentiss>
SocketCreateListener() failed -> probably a permission issue?
___nick___ has joined #freedesktop
<bentiss>
MrCooper: can you point at failing mesa jobs?
<MrCooper>
this might affect other projects as well though
<bentiss>
MrCooper: testing tright now the current pipeline on a runner that has no load whatsoever, and it's already failing :)
<daniels>
MrCooper: setting LP_NUM_THREADS=1 is probably also a good idea
<MrCooper>
hmm yeah, maybe it's the <number of test processes> * <number of llvmpipe threads> explosion?
<MrCooper>
still not clear why it's suddenly an issue now, not before
<MrCooper>
actually only (some of) the X server processes should spawn llvmpipe threads, not every test process
<bentiss>
that's weird. On that runner, with htop I don ´t even see the CPU getting used when everything explodes
<bentiss>
(without your patch)
<MrCooper>
it feels like some kind of artificial limit
<bentiss>
yeah, I just don't know where it comes from, and why it wasn't an issue before
<bentiss>
and OTOH, if it forces people to have a fair usage of CI, that's not so bad :)
<MrCooper>
true, but I'm worried that we can hit the limit even with reasonable utilization
<bentiss>
TBH, I am starting to wonder if we should keep debian testing as a base for those runners. Something like CoreOS or Silverblue (even flatcar linux if it weren't using docker) would be easier to maintain IMO
hikiko has joined #freedesktop
vbenes1 has joined #freedesktop
vbenes has quit [Read error: Connection reset by peer]
AbleBacon has quit [Read error: Connection reset by peer]
vbenes2 has joined #freedesktop
vbenes1 has quit [Read error: Connection reset by peer]
a-l-e has quit [Quit: Leaving]
hikiko has quit []
vbenes2 has quit []
vbenes has joined #freedesktop
phasta has quit [Ping timeout: 480 seconds]
phasta has joined #freedesktop
vladh has joined #freedesktop
<vladh>
I signed up for an account two hours ago but it's still "blocked", any tips?
<bentiss>
vladh: you should have received a verification email
<bentiss>
vladh:if not I can manually allow you in
<vladh>
bentiss: I haven't :( my email is vlad@vladh.net
<bentiss>
vladh: done
<vladh>
bentiss: thank you very much :)
<bentiss>
you're welcome (and sorry if it doesn't work quite well, it's an attempt at fighting bots)
pjakobsson has joined #freedesktop
<kbingham>
Aha... so signup is limited at the moment? My colleague djrscally also just tried to signup and hasn't been able to yet.
djrscally has joined #freedesktop
<bentiss>
kbingham: it's not. it's supposed to be working, with a delay. But it seems people who are using non standrad domains are having issues
<bentiss>
kbingham: if you need I can also unlock the situation for that person
<kbingham>
Ah. What's 'non-standard domain'? I expect he signed up with dan.scally@ideasonboard.com
<kbingham>
bentiss, Yes please!
<bentiss>
kbingham: approved
<kbingham>
bentiss, Thankyou
<bentiss>
kbingham: honestly no idea what would be non standard
<DavidHeidelberg[m]>
bentiss: oh you saved me! thanks
<bentiss>
make it harbor.freedesktop.org/cache/okias/ci-templates/ci-fairy:sha256-ef74372a6713cd93228278bacd7d6c8f09de0de0e50cc2ad8c9f61d92a45d35c in your case
<DavidHeidelberg[m]>
right, the harbor usage :) thx!
<bentiss>
well, you don't need to directly use harbor actually, the runners will pick it up
<bentiss>
so just registry.freedesktop.org/okias/ci-templates/ci-fairy:sha256-ef74372a6713cd93228278bacd7d6c8f09de0de0e50cc2ad8c9f61d92a45d35c as taken from your container registry
<bentiss>
oh, wait it might be my filter on the logs
<bentiss>
that was it. the POST log in the server doesn't give me s3.fd.o and I was filtering on it
<bentiss>
all good
<DavidHeidelberg[m]>
uffff
<DavidHeidelberg[m]>
I wonder, is any chance that session will decrease load on the servers? (there isn't new connection, but if it reuses existing, it could save a bit work).. just curious if it's relevant at this scale
<bentiss>
not sure I'll be able to measure that
<bentiss>
the thing is we have *a lot* of connections and they should use all 3 control planes, so hard to see a difference
<bentiss>
it would matter only at he nginx level too, I am not sure it keeps the same connection between nginx and ceph
<bentiss>
plus when you upload the file to ceph, it internally gets replicated over 3 nodes, so it might have some benefits, but I wouldn't bet anything on a reduce load of the cluster
<bentiss>
(the publish to quay happens when the pipeline is correct, at the very end)
<DavidHeidelberg[m]>
oh, thanks for merging :)
<MrCooper>
bentiss daniels: I wonder how many other projects use "ninja -j<...> test" in CI and expect it to limit the number of test processes spawned in parallel :/
<bentiss>
MrCooper: actually I should try to see if podman is capable of limitting how many cpus are exposed to the container
<bentiss>
I know it didn't work with docker
<MrCooper>
would be nice if it was possible
<bentiss>
IIRC there is gitlab-runner option that wasn't properly handled
<bentiss>
well, we can limit the cpu usage, not how many are exposed
<MrCooper>
taskset affects the nproc command, sadly not meson though
<bentiss>
given that it's not using cpu, but just spawning too many tests, it doesn't work to induce that quota
<bentiss>
(and the machine was loaded between 30 to 60 as some virglrenderer tests were also running)
<daniels>
last time I looked at that, we'd need to make podman use a runtime like Kata which would spawn a VM that we could limit the number of CPUs on
MrCooper has quit [Quit: Leaving]
MrCooper has joined #freedesktop
MrCooper has quit [Quit: Leaving]
MrCooper has joined #freedesktop
bl4ckb0ne has quit [Remote host closed the connection]
emersion has quit [Remote host closed the connection]
<anholt>
then you have to keep breaking down smaller and smaller sections to debug, when if you just had timestamps you could take your failing log and figure out where things stalled.
<eric_engestrom>
DavidHeidelberg[m]: too coarse
<anholt>
gitlab also managed to snatch failure from the jaws of victory by concluding that requests for timestamps were actually just requests for section timings.
<eric_engestrom>
(also, please don't use the phrase "make X great again")
<DavidHeidelberg[m]>
the quote is catchy, can't help it ;-) (for political reasons, I declare I don't care about politics)
<eric_engestrom>
oh, I was about to marge it (thanks for the acks!), but I just realized I need to bump container tags; just KERNEL_ROOTFS_TAG, right?
<anholt>
nothing in that MR affects containers
<anholt>
the CI *.{sh,txt,yml,toml} files go through artifacts
<eric_engestrom>
ah right
<eric_engestrom>
thanks :)
<eric_engestrom>
ok, marg'ing then
ybogdano has quit [Ping timeout: 480 seconds]
agd5f_ has quit []
agd5f has joined #freedesktop
Leopold___ has joined #freedesktop
Leopold_ has quit [Remote host closed the connection]
GNUmoon has quit [Remote host closed the connection]
GNUmoon has joined #freedesktop
Kayden has quit [Quit: to lunch and office for a bit]
ybogdano has joined #freedesktop
ybogdano has quit []
ybogdano has joined #freedesktop
___nick___ has quit [Ping timeout: 480 seconds]
mvlad has quit [Remote host closed the connection]
ybogdano has quit [Ping timeout: 480 seconds]
danvet has quit [Ping timeout: 480 seconds]
AbleBacon has joined #freedesktop
djrscally has joined #freedesktop
ybogdano has joined #freedesktop
GNUmoon has quit [Remote host closed the connection]