Investigate worker-deployment-related poll failure ergonomics #729
Description
Describe the solution you'd like
Today, as happened and is still happening in #691 despite it being closed, when a user hits a worker deployment limit, it fails poll calls which are just implicitly retried forever. Unfortunately, the ergonomics of worker deployments are such that they are implicitly created upon use, but require explicit deletion so the devex of knowing that once you get to your 101st deployment (100 is the current limit) things just stop working is bad.
Two things:
- See if there is any behavior improvement beyond the poll-failure warning (assuming that's occurring) to let users know they have hit deployment limit. But may be determined warning is all we can do because we don't want to consider this a fatal worker issue.
- Start deleting ephemeral worker deployments in cloud tests
Per-SDK Tickets
- Go -
- Java -
- Core -
- TypeScript -
- Python -
- .NET -
- Ruby -
- PHP -
- Temporal CLI -