The scope of this issue is fairly wide. I think it might need to be broken down a bit further.

Capture specific error points in terraform plan and terraform apply, especially around Docker socket access and provider authentication.

Which specific error points?

Log provider-specific authentication errors (e.g., AWS credentials) and expose them in both logs and UI.

We'll need to add code for each specific provider. How about we focus on the top 3: Docker, Kubernetes, AWS?

Add troubleshooting guides specific to template creation failures, covering Docker socket access, permissions setup, and provider authentication, with examples for Docker Compose, Kubernetes, and system services.

Are these not covered by our existing documentation?

• https://coder.com/docs/install/docker#permission-denied-while-trying-to-connect-to-the-docker-daemon-socket
• https://coder.com/docs/admin/integrations/multiple-kube-clusters

Are these not covered by our existing documentation?

They are but could be improved with explicit instructions (specifically regarding the docker socket permission issues in the context of Coder), also we need to link them from within the dashboard to take the user to actionable troubleshooting steps. I would prefer error codes and docs similar to the ones we have for the heathstatus page.

We'll need to add code for each specific provider. How about we focus on the top 3: Docker, Kubernetes, AWS?

Yes, We can start with these 3. It would be nice to have if we could run some preflight checks and suggest or auto-import the template the Coder server can connect.
e.g, If the docker socket is available and the Coder user has access to it, suggest using the docker template.

Which specific error points?

More verbose template build errors in the web template editor. They could relate to anything from wrong syntax to provider authentication issues.

Just got linked to this issue via support and this would be super helpful for users to be able to debug some of their own issues.

We're using Kubernetes and a workspace build with the wrong image tag (so it can't pull the image) or requesting more resources than can be scheduled leads to a completely opaque error. I've added some internal docs for how people should debug this (describe the deployment to see the events and then map those events error strings to what is actually wrong and how the fix it) but it's hard to make those docs discoverable and for people to know what to do.

If instead Coder could return those errors (either the raw error on the Kubernetes events for the deployment, or better yet something less raw and more user facing) then that would help a ton compared to just a time out waiting for the deployment to complete.

removing from docs for now - once we have it implemented or part of a release, we'll be able to document it