Enhance "Helm install failed: context deadline exceeded" errors

[craigwalton-dsit]

These are typically caused by:

• insufficient cluster capacity
• an error starting the container
• an error pulling the image

See docs https://k8s-sandbox.ai-safety-institute.org.uk/tips/troubleshooting/#im-seeing-helm-install-context-deadline-exceeded-errors

Oftentimes, the underlying error can be gleaned by viewing the Pod events for example in K9s (:pod, select relevant Pod, d, scroll to bottom). Should we try to extract these events to bundle into the error message, to make the user's life easier?

Some difficulties:

• By the time the helm install --wait fails, will Helm have automatically removed the resources?
• There can be many Pods in a release - do we infer which one is problematic based on what the status is?

[craigwalton-dsit]

I don't believe Helm will uninstall the chart if the installation fails unless the --atomic flag is set (which we don't currently do). So the resources should still be around for us to query.

I also tried using the --debug flag to helm install. This tells us which resources are not ready e.g.

ready.go:425: 2025-01-09 12:54:00.254703868 +0000 UTC m=+9.082961258 [debug] StatefulSet is not ready: craig-walton-default/agent-env-foo-default. 0 out of 1 expected pods are ready

but not why they're not ready, which can be cleaned from K9s/kubectl:

Failed to pull image "python:3.12-bookworm-doesnotexist": rpc error: code = NotFound desc = failed to pull and unpack image "docker.io/library/python:3.12-bookworm-doesnotexist": failed to resolve reference "docker.io/library/python:3.12-bookworm-doesnotexist": docker.io/library/python:3.12-bookworm-doesnotexist: not found