Rodrigo Campos

@tallclair I'm not sure if that would really help and it _might_ hurt for long lived connections or need to close connections going through that one? I can't remember, but...

@surajssd not sure I follow. There was some configuration that was working with one version, changed lokoctl to a newer version run `lokoctl cluster apply` and that directory is missing?...

@surajssd that is it to reproduce? No component for flex volumes nor anything? Cool, seems simpler to debug. Still very surprising, but let's see :)

@kolyshkin thanks for the cc! I'll be on and off for a few weeks. Is this something we want for the next runc release and I should try to prioritize...

@cheftako cool, good to know this approach looks fine if it is opt-in (as it is, yes :)). Thanks! And yeah, I'd like this to be simpler too. We will...

> Thanks for attacking this again. > > Do we really need more than one keystone container? Probably not for jobs, but when the _main_ functionality is achieved in conjuction...

> Precisely where my head was. We can "artificially" limit it to one for now > (and avoid defining semantics for > 1) and then lift that limit later. I...

@iamnoah thank you very much for your input too! I'll have a closer look at kubexit (I was not aware of it, it seems the first commit was in April...

/remove-lifecycle stale

The "features" sub-command will be nice to support to. It is becoming increasingly important for user namespace adoption to see if the OCI runtime support idmap mounts (via the features...