GNU bug report logs - #43079
guix ignores available substitutes

Previous Next

Full log

Message #11 received at 43079 <at> debbugs.gnu.org (full text, mbox):

From: Ludovic Courtès <ludo <at> gnu.org>
To: Jonathan Brielmaier <jonathan.brielmaier <at> web.de>
Cc: 43079 <at> debbugs.gnu.org
Subject: Re: bug#43079: guix ignores available substitutes
Date: Fri, 28 Aug 2020 16:23:31 +0200

Hallo!

Jonathan Brielmaier <jonathan.brielmaier <at> web.de> skribis:

> Tobias provided me some nginx logs from ci.guix.gnu.org so here is some
> further analysis:
>
> guix <at> laptop $ guix build --dry-run inkscape
> Folgende Ableitungen würden erstellt:
>    /gnu/store/i1sh5hzlzxzgzbdl3jb8wgnn9hr912wn-inkscape-1.0.drv
>    /gnu/store/1mzsr5z2qnrlr57gdv5mv7bhm1yg88z2-gtkspell3-3.0.10.drv
>    /gnu/store/07sff3s4ckc81j2lxip0pcyikfqbqs6v-gtkspell3-3.0.10.tar.xz.drv
>
> Where the inkscape drv derives to
> /gnu/store/wp4v7qsyjngxggncz34gaidb3yqcqy3l-inkscape-1.0
>
> https.access.log:XXX.XXX.XXX.XXX - - [28/Aug/2020:00:48:08 +0200] "GET
> /wp4v7qsyjngxggncz34gaidb3yqcqy3l.narinfo HTTP/1.1" 200 3336 "-" "GNU Guile"
>
> It's also in the cache:
> -rw------- 1 root root 3451 28. Aug 01:18
> /var/guix/substitute/cache/4refhwxbjmeua2kwg2nmzhv4dg4d3dorpjefq7kiciw2pfhaf26a/wp4v7qsyjngxggncz34gaidb3yqcqy3l
>
> (narinfo (version 2) (cache-uri "https://ci.guix.gnu.org") (date
> 1598570321) (ttl 7776000) (value "StorePath:
> /gnu/store/wp4v7qsyjngxggncz34gaidb3yqcqy3l-inkscape-1.0\nURL:
> nar/gzip/wp4v7qsyjngxggncz34gaidb3yqcqy3l-inkscape-1.0\nCompression:
> gzip\nFileSize: 40257671\nURL:
> nar/lzip/wp4v7qsyjngxggncz34gaidb3yqcqy3l-inkscape-1.0\nCompression:
> lzip\nFileSize: 16181383\nNarHash:
> sha256:1i44f21lnxmajdir2jqvg9d28518fipbv3jq4cpf6wmmk1xgqr8m\nNarSize:
> 164233984\nReferences: 01b4w3m6mp55y531kyi1g8shh722kwqm-gcc-7.5.0-lib
> 07m4rm6cwdj2m5mzh6n19wldn1if3h0q-glibmm-2.62.0
> [...]
> zrmhnj3pwchn2msphgnwzwd3q89m46rn-aspell-0.60.8\nSystem:
> x86_64-linux\nDeriver:
> s8f2r8md9sjmmgnac1k5y881d6a84c90-inkscape-1.0.drv\nSignature:
> 1;berlin.guix.gnu.org;LONG_KEY\n"))
>
> I have no idea what is wrong here...

I’m not sure what conclusions you’re drawing here?  :-)

As you found, each entry in /var/guix/substitute/cache has a TTL,
including for negative lookups.  This is why one can observe different
behaviors on different machines: one machine can think the substitute is
unavailable (cached negative entry not yet expired), while the other got
a positive lookup soon after the substitute had been “baked” on the
server.

TTLs vary.  For successful lookups, this is usually a long TTL (see
‘guix publish --ttl’).  For negative lookups, there are two cases: a
1h-or-so TTL for “absolute no”, and a 5mn TTL for
“substitute-being-baked no”.

On top of that, nginx on ci.guix.gnu.org might perform extra caching,
though we’d need to check the config.

HTH!

Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.