Worker Versioning -- Getting Started

Developer Corner
, ,
1

The instructions are a bit unclear to me — after making changes to workflows, do I need to redeploy my workers with the updated configuration parameters?
For example, if I change the build_id (which we bump whenever workflows are updated to trigger versioning behavior, i suppose), do I also need to redeploy the workers via CLI with those new parameters?

In our case, we have long-running workflows (that stay active indefinitely). I tried deploying the worker changes in the staging environment without the CLI changes. In Cloud Temporal, it shows that the workers were deployed with the new build_id, but they don’t pick up any activities. When I check the call stack or run queries, I get something along the lines of "no workers are polling”.

Interestingly, when I run the same setup locally to test, everything works fine, and the behaviour of AUTO-UPGRADE works with no issues at all.

2

Hi

yes, yes you will need to redeploy the workers with the new build id and instruct temporal (using temporal cli) to route the tasks to the new workers(Build ID)

In Cloud Temporal, it shows that the workers were deployed with the new build_id , but they don’t pick up any activities. When I check the call stack or run queries, I get something along the lines of "no workers are polling”.

Please open a ticket https://temporalsupport.zendesk.com/ if you are having issues.

3

Hi Antonio,

Thanks for the quick reply. I’d like to confirm my understanding of the workflow change:

Previous behavior (non worker versioning):

  • When we redeployed a new worker, it would immediately start processing activities without requiring any Temporal CLI commands

Current behavior (worker versioning):

  • When we deploy a new worker due to some workflow changes, we need to use the Temporal CLI to route tasks to the new workers using the Build ID

  • A deployment_name, build_id parameter is now required for worker deployment (reference)

Is this understanding correct?

Regarding our staging issue: Could this new deployment_name, build_id requirement be related to why our workers in staging appear as deployed with the new build_id but aren’t picking up activities (showing “no workers are polling”)?