<gchristensen>
ok andi-, I'm seeing in the build log of an Aborted build that it actually Timed Out... but for whatever reason Hydra isn't getting the message.
<andi->
so we now have the other edge case for timeouts? Lower then the 10h we thought it would be?
<gchristensen>
looks like the timeout it is tripping on is 4hrs, looking in to why ...
<gchristensen>
oh well that'll do it
<andi->
gchristensen: thank you, feeling terrible not being able to help with that. :)
<gchristensen>
ok andi-, packet-t2a-1 had buildCores set to 0, -2 had it set to 1. I'm going to set both to 4. sgty?
* andi-
wonders about that 0
<andi->
sounds better at least :)
<gchristensen>
cores=0 means use however many
<andi->
yeah that sounds like it could lead to trouble
<andi->
lets see if things improve
<gchristensen>
we'll make them both have similar configs and then we can do Science by changing only one node
<gchristensen>
-1 and -2 are now running with 4 cores.
<niksnut>
aminechikhaoui: I think we need to revert (or disable by default) the use of TransferManager, it's causing frequent hydra-queue-runner aborts: https://github.com/NixOS/nix/issues/2493
<{^_^}>
nix#2493 (by edolstra, 2 days ago, open): Unhandled exception in AWS TransferManager thread
<andi->
gchristensen: nice, you annotated the graph. That just popped up here.
<gchristensen>
oh, cool!
<gchristensen>
it feels surprising that AWS's TransferManager is doing the wrong thing
orivej has joined #nixos-dev
pie__ has joined #nixos-dev
pie_ has quit [Ping timeout: 250 seconds]
<gchristensen>
andi-: reload and look at the pink one?
<andi->
as soon a firefox unfreezes o.O
<gchristensen>
I broke it anyway
<gchristensen>
ok fixed
<andi->
I only see packet-t2a-2:9100 in there, did you start another exporter there?
<gchristensen>
ehh for some reason grafana isn't showing the annotation for both nodes which were deployed at the same-ish time
kgz is now known as kgspooks
<gchristensen>
globin, LnL -- maybe y'all have advice?
<{^_^}>
#49427 (by mztriz, 12 hours ago, open): make-system-tarball.sh: double quote to prevent globbing and word splitting.
<Mic92>
also I don't know what if the event was the motivation for that.
Sigyn has quit [*.net *.split]
<Synthetica>
Oh well, itll all be over the day after tomorrow. For next year: can't we ask DigitalOcean to blacklist nixpkgs in the list of popular repos?
<Mic92>
Atm the spam is easy to filter.
<Mic92>
I think it also motivate some other people to publish their fixes.
<arianvp>
gchristensen: could you perhaps add me to ofborg?
<arianvp>
I've opened a PR :)
<globin>
gchristensen: is there still an issue? if yes I'm not sure what you mean?
<gchristensen>
hrm, I'm surprised ofborg doesn't catch that
Sigyn has joined #nixos-dev
<gchristensen>
ok so the two aarch machines have stabilized at about 50% cpu -- an improvement over the 100% pegged cpu :)
<andi->
now that aborting issues has to go away and then things will be beautiful (again) :-)
<gchristensen>
also eelco has fixed a couple bugs in Nix
<andi->
I just saw that thing with the transfer manager change
orivej has joined #nixos-dev
<aminechikhaoui>
niksnut: yeah sure, but looks like you fixed it already ?
<aminechikhaoui>
btw I didn't deploy that yet to our hydra :D it was a one time thing
<aminechikhaoui>
oh you just made it configurable
<niksnut>
yes
drakonis1 has joined #nixos-dev
xeji has joined #nixos-dev
<Profpatsch>
TIL: openjdk has a dependency on polkit, and polkit has a dependency on spidermonkey.
<Profpatsch>
Both Runtime of course.
<domenkozar>
all java roads lead to script
<clever>
Profpatsch: ive looked into the spidermonnnnnnnnnnnnejy stuff before, and policykit uses "javascript" to define the policies, and then runs it under spidermonkey to evaluate the policy
<drakonis1>
polkit is the worst
<gchristensen>
I'm going to break down and add the MLX NIC configs to the linux kernel options.
<gchristensen>
I don't like it, but also waiting for a kernel compile vs. shaving a second off hydra... to deploy hydra... is silly
<gchristensen>
# Then do test your change by building all the kernels (or at least# their configs) in Nixpkgs or else you will guarantee lots and lots# of pain to users trying to switch to an older kernel because of some# hardware problems with a new one. <-- nice
<Synthetica>
Is there a reason Cuda isn't being updated?
<Mic92>
gchristensen: nix-review?
<gchristensen>
hrm?
<Mic92>
gchristensen: to test all kernels. I usally have also some remote workers to distribute the load.