Or should we use the “Wait for a Machine to Reach a Specified State” API right after calling it.
In case the start machine API does wait before returning, is there a timeout as well? Is it possible that a stopped machine will not be able to start because no allocation is available any more?
The start endpoint waits until the VM init is responsive so you don’t need to use the wait endpoint.
Yes, I believe there’s a max 60 second timeout on the endpoint.
Yes, it is possible there’s not enough capacity on the host based on the resources for the machine. When that happens, it errors pretty quick and records a failed event on the machine with the reason.
When an allocation does not happen, what status does the machine get into?
It gets transitioned to failed and the event data shows why.
Cool. In that case, the docs can also be updated with the failed status here.