Increase timeouts on CI

[apostasie]

Either GH is busier or they downgraded the base instances we are using.

One way or the other:

• windows unit test now hit the 5 minutes mark
• building dependencies on non arm machine hit over 10 minutes mark
• ipv6 tests are hitting the 10 minutes mark

@AkihiroSuda @djdongjin at your convenience.

[djdongjin]

LGTM, thanks

[apostasie]

Thanks @djdongjin

Just increased ipv6 test as well.

Not sure if GH slowness is transient or not, but either way we need a bit of leeway and I did set these timeouts too aggressive in the first place.

[apostasie]

Ok... something is very wrong with GH right now...

https://github.com/containerd/nerdctl/actions/runs/12150865305

[apostasie]

@djdongjin there definitely was some transient issue with github causing a bunch of timeouts.

I still think relaxing the timeouts a bit on these is warranted and will not hurt anyhow (windows normally clocks in at 4min+, quite close to the current 5 mins limit), so, let's merge if you are ok with that.

[fahedouch]

is the execution engine (e.g. Docker) that runs these jobs using the maximum CPU/memory available on the machine ? we will reach a limit with the timeout :/

[apostasie]

is the execution engine (e.g. Docker) that runs these jobs using the maximum CPU/memory available on the machine ? we will reach a limit with the timeout :/

Not sure. I will have a look around.

About that, I have mixed feelings about the fact that we are running tests inside a docker container overall:

• this does introduce a layer of complexity (and possible additional bugs - especially on networking, but maybe also snapshotter?)
• this also alters the type of situations we normally see on a host, by restricting us to this "simplified" docker env

I know we do that for a reason (provide environment consistency and reproducibility), but it comes at a cost ^.
Maybe we could have canary run on the host instead of inside the container (would obviously require a lot of work to set the env though, so...).

[apostasie]

@fahedouch

https://github.com/containerd/nerdctl/pull/3726/checks#step:3:87

docker info reports 4 CPU 16GB.

Is there somewhere else dockerd could constrain containers resources that would not show in docker info?