Jacob Blain Christen

What if SUC took existing labels+annotations that aren't in a `cattle.io` "namespace" from the Plan and set them on the generated Job and Pod template? Are there use-cases where that...

Is this a problem with SUC or a problem with assumptions baked into https://github.com/k3s-io/k3s-upgrade/blob/a78cfaf2446b8d6b014a76c2812a1349e1738e37/scripts/upgrade.sh ? Likely more appropriate to bring this up with the k3s project directly: https://github.com/k3s-io/k3s/issues

As upgrades are implemented leveraging SUC, this suggestion would be more appropriate there: https://github.com/rancher/system-upgrade-controller/issues

Shoot, this smells like a wrangler bug. Or maybe docker/distribution, oh no.

@brandond Can you try with an image with a path deeper than two? Should be supported by default with `registry:2`, e.g. `/this/is/one/long/but/legit/name:tag`

> @brandond Can you try with an image with a path deeper than two? Should be supported by default with `registry:2`, e.g. `/this/is/one/long/but/legit/name:tag` I suspect a bug in either `rancher/wrangler`...

It looks like this is getting a zero byte direntry: https://github.com/diskfs/go-diskfs/blob/master/filesystem/iso9660/directoryentry.go#L279 (is proxmox doing something goofy? can you mount and inspect the cdrom geing generated?)

I do not understand why removing the user constraints put on a container for `docker.inside` is under consideration. The `docker.inside` mechanism is for injecting a workspace into a container so...

@antoniobeyah wrote: > Will this PR break your environment somehow? If so I may be able to add in some checks to make that less likely, just need some specific...

@antoniobeyah @jglick: My other qualm with this is that I do not want my build containers to run as root unless I explicitly tell them to via code (as that...