[Nix-dev] Intermittent failures for containers test

[Mateusz Kowalczyk]

Hi,

Often I look on Hydra to see why nixos-unstable channel hasn't moved and
often I see nixos:trunk-combined:nixos.tests.containers as the reason.
Looking at the log and from an IRC discussion I had few days ago, the
problem is stipulated to be networking not coming up quickly enough. Log
at [1].

In cointainers.nix there is a:

#$machine->succeed("ping -c1 $ip"); # FIXME

This is commented out, probably because it is a workaround. Is anyone
actively working on finding the bigger problem, and if not, can we
perhaps uncomment the line? It's a shame to not get channel updates due
to an intermittent failure which doesn't reflect actual changes in
nixpkgs: it's not testing whether we made changes which makes
containers.nix test fail as much as it tests our luck.

Of course, this workaround might just not work or run until the job
times out at which point it is less useful but would benefit from
further documentation rather than just ‘FIXME’.

It'd just be great if someone could shed some like on the problem here.

[1]: http://hydra.nixos.org/build/13049666/log/raw
-- 
Mateusz K.