[Cuirass] Queued derivations have their GC roots removed

  • Done
  • quality assurance status badge
Details
2 participants
  • Ludovic Courtès
  • Ludovic Courtès
Owner
unassigned
Submitted by
Ludovic Courtès
Severity
normal
L
L
Ludovic Courtès wrote on 20 Oct 2024 22:06
(address . bug-guix@gnu.org)
87cyjuk088.fsf@inria.fr
Cuirass stores GC roots for all the derivations of its jobsets under
/var/guix/profiles/per-user/cuirass/cuirass (or similar).

The ‘delete-old-gc-roots’ periodically deletes these GC roots (which are
not just derivations), but it’s supposed to keep GC roots to derivations
that are queued: that was done in
55af0f70c0d4938b8eda777382bbc4d8f5698a37.

Unfortunately, there’s ample evidence on berlin that GC roots of .drv
currently queued get removed anyway. This causes workers to try in vain
to substitute them, to eventually reject the build, which gets canceled
by ‘remote-server’.

Ludo’.
L
L
Ludovic Courtès wrote on 10 Nov 2024 12:49
(address . 73916@debbugs.gnu.org)
87frnzwbok.fsf@gnu.org
Ludovic Courtès <ludovic.courtes@inria.fr> skribis:

Toggle quote (13 lines)
> Cuirass stores GC roots for all the derivations of its jobsets under
> /var/guix/profiles/per-user/cuirass/cuirass (or similar).
>
> The ‘delete-old-gc-roots’ periodically deletes these GC roots (which are
> not just derivations), but it’s supposed to keep GC roots to derivations
> that are queued: that was done in
> 55af0f70c0d4938b8eda777382bbc4d8f5698a37.
>
> Unfortunately, there’s ample evidence on berlin that GC roots of .drv
> currently queued get removed anyway. This causes workers to try in vain
> to substitute them, to eventually reject the build, which gets canceled
> by ‘remote-server’.

This was fixed or at least mitigated by these Cuirass commits:

e96f088 * register: Add ‘--derivation-ttl’ option, default to twice ‘--ttl’.
304c09b * base: Be more conservative in deciding whether to remove a .drv GC root.

In particular, .drv for builds currently in ‘failed-derivation’ status
are no longer GC’d when other builds in that evaluation might eventually
trigger a failed-derivation -> scheduled transition.

Ludo’.
L
L
Ludovic Courtès wrote on 10 Nov 2024 12:50
control message for bug #73916
(address . control@debbugs.gnu.org)
87ed3jwbo6.fsf@gnu.org
close 73916
quit
?
Your comment

This issue is archived.

To comment on this conversation send an email to 73916@debbugs.gnu.org

To respond to this issue using the mumi CLI, first switch to it
mumi current 73916
Then, you may apply the latest patchset in this issue (with sign off)
mumi am -- -s
Or, compose a reply to this issue
mumi compose
Or, send patches to this issue
mumi send-email *.patch