flyctl

flyctl copied to clipboard

We could allow trust checks by running a custom ephemeral machine, where if the machine returns 0 the trust check passes, and otherwise it failed.

This would be useful for doing more complex health checks, like making sure signup functionality continues to work after a deployment.

@billyb2 Can you expand on what the workflow for this would look like? Is it something we could use fly m run for instead?

I just mean that basically, flyctl after deploying the machine but before marking it as available (or uncordoning or whatever), spawns a new machine with fly m run --autodestroy. If that machine exits and returns 0, then the machine passed the health check. Otherwise, it didn't. This is especially useful for things like bluegreen deploys.