<jtojnar>
vcunat: could you return the gnome jobset back to gnome-3.26 branch, please?
<vcunat>
jtojnar: OK, let me restart it
<LnL>
could we make hydra block the release channel?
<vcunat>
LnL: what do you mean?
<LnL>
the hydra build broke again on 17.09
<vcunat>
force it not to update the channel?
<LnL>
so I can't update without hydra.package = { outPath = /nix/store/...; }
<vcunat>
oh, you mean adding the "hydra" job to the set that must succeed
<LnL>
yeah include the package/test in the jobset so the channel doesn't update
<vcunat>
heh, due to nix update
<vcunat>
sounds OK to me, but why was it removed?
<vcunat>
(you seemed to imply that)
<vcunat>
jtojnar: error: getting status of '/nix/store/ing9v5d9dfd0dc7n1afzk095riy7yvap-git-export/pkgs/desktops/gnome-3/apps/bijiben/no-update-icon-cache.patch': No such file or directory
olejorgenb[m] has quit [(Ping timeout: 248 seconds)]
pstn has quit [(Ping timeout: 241 seconds)]
moredread[m] has quit [(Ping timeout: 264 seconds)]
hedning[m] has quit [(Ping timeout: 255 seconds)]
copumpkin has quit [(Ping timeout: 255 seconds)]
florianjacob has quit [(Ping timeout: 240 seconds)]
nocent has quit [(Ping timeout: 250 seconds)]
grahamc has quit [(Ping timeout: 248 seconds)]
florianjacob has joined joined #nixos-dev
FRidh has quit [(Ping timeout: 240 seconds)]
FRidh has joined joined #nixos-dev
__Sander__ has joined joined #nixos-dev
stites[m] has joined joined #nixos-dev
regnat[m] has joined joined #nixos-dev
nocent has joined joined #nixos-dev
hedning[m] has joined joined #nixos-dev
pstn has joined joined #nixos-dev
grahamc has joined joined #nixos-dev
sphalerite has joined joined #nixos-dev
olejorgenb[m] has joined joined #nixos-dev
hl has joined joined #nixos-dev
copumpkin has joined joined #nixos-dev
rycee has joined joined #nixos-dev
moredread[m] has joined joined #nixos-dev
FRidh[m] has joined joined #nixos-dev
peterhoeg has joined joined #nixos-dev
FRidh has quit [(Ping timeout: 240 seconds)]
orivej_ has quit [(Ping timeout: 268 seconds)]
FRidh has joined joined #nixos-dev
orivej has joined joined #nixos-dev
<vcunat>
LnL: I suspect that there are a few jobs stuck for days in "building state" - and all others currently depend on them and wait (~1000 jobs)
<vcunat>
that together paralyzes the 17.09-darwin channel
<vcunat>
niksnut: ^^
<vcunat>
(I can only restart jobs etc. and that seems not to help at all. Restarting cancelled jobs that still build just seems to resume waiting on them.)
<bgamari>
does anyone know how one might set crossSystem and build a nixos image?
<mguex>
vcunat: nice. the distinction is that we want to dedirect requests from cache.nixos.community to mirrors that have the requested nar file (e.g. the http server of your local university)
<copumpkin>
vcunat: thanks
<mguex>
cache.nixos.community (or whatever domain we actually chose) gets 3-5 A/AAAA records with a short TTL so we can switch to another host if one goes down
<mguex>
the answering host looks up the requested file in the redis db and redirects the client to a mirror close to her (i.e. same ASN)
<vcunat>
mguex: I can't really see the advantage over fetching from IPFS directly
<mguex>
vcunat: there is not an advantage
<mguex>
however, my tests showed that IPFS has a lot of overhead currently, and you need a rather beefy machine for it
<mguex>
Currently you cannot just sync 350 GiB with acceptable data rates - it would take a long time
<mguex>
you actually need to shut the daemon down, bring it up in "stealth" mode and connect to other nodes manually to do sync with data rates @ ~ 10 Mb/s
<vcunat>
mguex: so the fetch over IPFS is only capable of 10 Mb/s ? (I'm not sure I understood this.)
<mguex>
yes
<mguex>
and that is with all tricks (the mentioned stealth mode)
<mguex>
without that it's way slower
<mguex>
that has to do with the constant DHT management traffic/updates
<mguex>
I hope they will find a solution (maybe with their 200 mi.$ ICO)
<mguex>
I realized far to late that IPFS is called pied piper (look up the meaning if you don't know it) in the U.S. series "Silicon Valley" :P
<copumpkin>
niksnut: you hear that AWS is moving towards KVM backend away from Xen? or at least appears to
<copumpkin>
need to make sure we get those virtio drivers into the AMIs :D
<copumpkin>
although I think we already have them there
<niksnut>
copumpkin: interesting
<niksnut>
I guess we'll want virtio in the initrd
<copumpkin>
it's only on the new C5 instance types for now
<copumpkin>
but apparently they plan on moving more stuff over there
<niksnut>
heh, the current spot price for c5.18xlarge is significantly below c5.9xlarge and c4.8xlarge at the moment in eu-west-1
<vcunat>
mguex: hmm, unfortunate. Maybe IPFS could still be leveraged to distribute meta-information, but it might be easier to choose a completely different solution. There are e.g. some open-source SW for managing mirrors.
<niksnut>
maybe I'll do some benchmarks tonight
<copumpkin>
:D
<mguex>
vcunat: yeah, fpletz installed mirrorbits, it just needs a HA redis setup (sentinel)
<mguex>
vcunat: I believe that IPFS can still be used - but with the current manpower behind the mirror project, rsync seemed to be the better solution for now
<mguex>
a lot of scripting needs to be done to make everything perform with acceptable rates
<copumpkin>
it's the thing we mentioned at the hackathon
<copumpkin>
hmm I seem to have an extraneous comment in there
<copumpkin>
not sure how that crept in
<copumpkin>
I'll fix later I guess :)
<niksnut>
what does (literal "/tmp") (subpath TMPDIR) mean?
FRidh has quit [(Remote host closed the connection)]
<mguex>
vcunat: Like you need to notify all pinners through some channel (ipfs's pub-sub, irc, xmpp, http) so that they bring down their "public" IPFS daemon, start in stealth mode, do the sync and come back online (dht mode)
<mguex>
if you do not want to notify, you can also check whether the /ipns/ pointer entry has changed, if yes do the sync thing
zraexy has quit [(Ping timeout: 250 seconds)]
zraexy has joined joined #nixos-dev
FRidh has joined joined #nixos-dev
<copumpkin>
niksnut: means you can hit exactly the /tmp path and any subpath of the TMPDIR folder
<copumpkin>
the /tmp thing is just to find out that it's a symlink
<copumpkin>
TMPDIR is passed in from the outside as a parameter
<niksnut>
oh yeah I remember
JosW has joined joined #nixos-dev
Sonarpulse has joined joined #nixos-dev
Sonarpulse has quit [(Remote host closed the connection)]
<jtojnar>
btw what is the relationship between grahamcofborg and nixborg?
<gchristensen>
grahamcofborg is a community-driven PR tester, nixborg is to be more official and have more thoroughtesting
<MoreTea>
pierron, (or someone else familiar with the internals of the module system): would you have any suggestions on how to get the meta.maintainers from modules?
<MoreTea>
I'm able to compute them for lots of packages and the nixos tests, but how to get this information for modules eludes me at the moment
jushur has quit [(Quit: The Borg joined forces with Skynet, Resistance is futile! Uploading has begun!)]
orivej has quit [(Ping timeout: 248 seconds)]
_ts_ has quit [(Ping timeout: 240 seconds)]
<pierron>
MoreTea: that should be easy to do, just open an issue and cc me on it