<gchristensen>
starting to prefer my "Delete it all" strategy :D
<simpson>
Guess even Azure can embrace and extend~
<simpson>
I'm a little confused as to why Azure gets special treatment here in the code. It seems like the other backends don't have this extra snowflaky structure.
<gchristensen>
I think b/c it isn't clear the right way to do things
<gchristensen>
afaict that logic should be partof the azure nix module, not the eval machine info
<simpson>
A large part of nixops' existence is being a weirdness containment facility.
<gchristensen>
I mean just the unconantained weirdness :)
<simpson>
Oh, well IIRC Azure support was written by one person under GSoC or similar. In that case, it's just as much our fault for failing to help them integrate.
<simpson>
What you said before: The right way to add a new vendor/backend isn't documented or illustrated.
<{^_^}>
nixops#981 (by grahamc, open): Delete all the providers (and use plugins instead)
<simpson>
gchristensen: I'd like to read the code that loads plugins; where should I look in this 216-file change?
<simpson>
Also, opinion: Python plugin technology is all shit; you should use straight.plugin or bravo.plugin but be prepared for sadness. Therefore, while I like the concept of a looser coupling, I think that all of the code should stay in a single repository, just with a more aggressive wall between them.
<gchristensen>
simpson: "Note, for a better view of the diff -- check this out:"
<simpson>
OTOH it could be that that was the original intent and we're just not achieving it.
<simpson>
You use NIE to indicate incomplete API, but that's an ad-hoc solution; instead, plugins should publish the list of things that they can do.
<simpson>
I have returned to a day job working on Django monoliths. I need to find inner peace and not be angry at Python.
<gchristensen>
"ask for forgiveness, not permission" -- python for some reason :P
orivej has quit [Ping timeout: 244 seconds]
<FRidh>
not specifying specific exceptions...how nice
<FRidh>
reminds me of code at work :(
vcunat has quit [Ping timeout: 256 seconds]
<andi->
the people that seem to have an "issue" with python seem to be gathering here.. we could all hold hands and hope for better times :)
vcunat has joined #nixos-dev
goibhniu has joined #nixos-dev
FRidh has quit [Quit: Konversation terminated!]
<vcunat>
niksnut: t2a.cunat.cz should be ready now. I did basically the same setup, though just 4GiB RAM, disabled swap and tmpOnTmpfs in this case, as it seemed enough. The "metrics" build peaks at 3 GiB, so we'll have a hard failure if we regress too much - that's a feature and not a bug :-)
<{^_^}>
vcunat: 23 hours, 39 minutes ago <gchristensen> please ping me (gchristensen) when you're around, w.r.t. your cache slowdown. Please run `curl -I https://cache.nixos.org` a few times, and send me the results (headers specifically)
<vcunat>
Oh, I see my ping to cache.nixos.org is 4-5 ms on that machine again. Thanks aminechikhaoui.
<vcunat>
(well I hope it will last long)
<gchristensen>
yay!
<gchristensen>
vcunat: I suggested re-running b/c I added header capture to the script, not jst b/c of the -4 option
<vcunat>
Apparently the information was enough for them to fix it.
<gchristensen>
they wanted additional traces just in case, but yes
<gchristensen>
(my requests of you were requests from AWS)
<gchristensen>
in fact, I think they would still like them
<vcunat>
I don't have any nixified machine online on that location ATM, so I can send them on Monday (if I remember).
<vcunat>
(I just checked ping quickly from a router.)
<gchristensen>
I don't think it needs to come from a nix machine, just any machine
<gchristensen>
also, it can wait
<vcunat>
gchristensen: right, after tweaking the shebangs I got http://ix.io/1iB6
<vcunat>
there's no `traceroute -6` support on the router, apparently
<gchristensen>
thanks!
<vcunat>
(it's just a cut-down busybox implementation)
pie_ has joined #nixos-dev
<gchristensen>
nice
<gchristensen>
fwiw this is what they had to say: " That being said I see this user was routed to to an Edge location in the North America (South Bend).In my tests against your distribution I am also routed to this Edge location."
orivej has joined #nixos-dev
pie_ has quit [Remote host closed the connection]
pie_ has joined #nixos-dev
phreedom has joined #nixos-dev
phreedom_ has quit [Ping timeout: 250 seconds]
goibhniu has quit [Ping timeout: 248 seconds]
phreedom has quit [Ping timeout: 250 seconds]
pie_ has quit [Ping timeout: 244 seconds]
phreedom has joined #nixos-dev
<niksnut>
vcunat: thanks, I added the machine, but I didn't realize it's a darwin machine
<niksnut>
so that won't help us since the metrics jobs are x86_64-linux
<gchristensen>
niksnut: he's provided 2 machines I think
<samueldr>
it's not detailed yet, and I'll be working on things even if the RFC doesn't receive any comments nor approval
<samueldr>
part of the RFC is from the name "Request for Comments", I'll be requesting comments :)
<andi->
My initial feelings are: Process documentation and guidelines => Yay! A team for that => not so sure but maybe people that can be used to give advice/comment
<andi->
I'll think about that more before giving feedback there
<samueldr>
yeah, and wait for the details, the team may be more fluid than you think
<andi->
samueldr: anyways. Thank you for working on it!
<andi->
samueldr++
<{^_^}>
samueldr's karma got increased to 9
<samueldr>
I spent a couple hours thinking notes up about processes to **ensure no burnout**