Leaving this one in Community MVP, but removing from switch blocker. This is an issue that causes confusion and hurts UX and resolving this will make it easier to debug and troubleshoot failed provisions, but is also a nice to have for flipping the switch to public in May.

@tjcran have you explored potential solutions to this problem? Maybe there's something simple we can do that makes it a lot better.

Keep in mind this doesn't exist in v1 since once you get the infrastructure working, the configs don't change so it will probably keep working.

Does being able to dump the Terraform template/state from the CLI seem sufficient for this, or is there more to do here?

FWIW, I also recently ran into some issues with my dev environment where this would have been helpful.

@dwahler State push/pull is a necessary addition to the product, but we have to be careful about leaking secrets.

Good point. Maybe something that's equivalent to terraform state show would be better, since that redacts variables that are marked as sensitive.

Good point. Maybe something that's equivalent to terraform state show would be better, since that redacts variables that are marked as sensitive.

We do have coder state pull <workspace> which helps. For example:

$ coder state pull ab 2>&1 | jq -r '.resources[] | select(.type=="kubernetes_deployment").instances[0].attributes
.metadata[0].name'

pulls the deployment name for my workspace. This is cumbersome, but perhaps sufficient.

Setting an agent connection timeout through terraform was a suggestion by @kylecarbs in grooming today.

Wait for the agent to be up since a user cannot do anything anyway until they are connected was a suggestion from @dwahler

The timeout should be set via our Terraform provider. Here's an example:

resource "coder-agent" "dev" {
  timeout = "1m"
}

Upon SSH, the user could be notified with a message that the agent timed out which isn't perfect, but is better than spinning forever.

Moving this to an Epic after our grooming discussion. Tickets needed:

Time out on agent connect after a successful build (2 min, 10 min?). This one needs fixed before Community MVP launch.

• Add CLI command and Show the agent connect failure on the FE for users. Potentially dump all the names of the resources used. Add meta data (key-value mapping related to a resource) instead and an FE ticket to show this information to the user. See ticket Get infrastructure link for each resource #1325
• Error when build fails - The error property is Exit 1 - make this more useful.