A. Presumably binderhub should take a different response then keep trying forever if it can’t get the image.
B. If you have a way to check for images that don’t work (cache says they should exist but they don’t), it would be nice if such images could get kicked from the cache.
Sadly no. I gave up after I went to gesis where it worked (albeit extremely slowly). I got that one to work by going directly to gesis.
How would I tell which backend member I was on, would a copy of the build log have that info?
P.S. This wasn’t the first time it had failed like this. It had done this several times over a few weeks, so if you tell me how to find this I’ll look to find out what backend is failing if this should happen again.
If you can copy the build logs when it fails, that will help. I did find a stuck build from your repo on our GKE cluster, so I deleted that. It seemed to build and launch promptly on OVH. It may be due to one or more unhealthy nodes, which can cause slow launch times due to disk pressure or other issues.