The described scenario of Tableau Server being unable to start is expected when one of the workers is unreachable.During start-up the primary Tableau Server must be able to reach the worker machines to setup initialization, synchronize the data engine files and synchronize the repository. I tested in both 8.3 and in 8.2 branches and the behavior is the same.
The failed start-up message allows administrators to be notified when one of their machines is not behaving as expected at start-up and recover the machine before starting the cluster.
Rest assured that if one of the servers goes down after startup while the server cluster is running, the Tableau high availability service will switch over to the other server and recover on its own.
It is understandable that since the worker server contains only redundant elements the cluster could conceivably be able to start without the worker but that is not a feature currently built into the product.
If you would like to submit that as a feature request, please create an Idea post in our Ideas forum: Ideas
I hope this helps!
And now that enhancement request is in "ARCHIVED" state, which probably means nothing will be done about it.
Just curious, are you still using version 8.3.0?
We’re currently on 10.0.3