Using Replay before deploying a new Workflow version with Temporal

When making code changes to a Workflow, it is important to avoid introducing non-deterministic behavior.

Patrick Rachford
3 min readOct 24, 2023
Replayer code in TypeScript

As you iterate and refine your codebase, one of the most significant challenges you might face is preventing the introduction of non-deterministic behavior. A solution? Temporal’s Workflow Replayer. This tool serves as a safety net, allowing developers to pre-emptively test and validate their changes against existing Workflow Histories before a production deployment.

One way to prevent this a non-deterministic error from occurring in your long-running Workflows, is to use the Workflow Replayer to replay existing Workflow Histories against your new code before deploying it to production.

While the Workflow Replay helps to avoid non-deterministic code, it does not guarantee that your code is deterministic.

How to use the Workflow Replayer

To use the Workflow Replayer, you need to provide the following:

  • The Workflow Replayer to register the Workflow that you want to replay
  • A Workflow History to execute, either:
  • Loaded directly from the Client
  • Loaded from the JSON History file

Consult the SDK specific documentation for information on using the Workflow Replayer programmatically.

If you chose to load the Event History from a JSON file on disk, you can do so programmatically or from one of the following methods:

Using the Temporal CLI

Use the Temporal CLI to download the Workflow History file from the CLI:

temporal workflow show - workflow-id <workflow_id> - output json > <output_file>

Using the Temporal Web UI

Navigate through the Temporal Web UI to download the Workflow History file:

  1. Navigate to the Workflow Execution you want to download the History from.
  2. In the Recent Events section, select Download.

Once you have the Workflow JSON History file, you can use the Workflow Replayer to replay the Workflow History, in the SDK of your choice.

Replay existing Workflow Histories in CI

Use the Workflow Replayer in your existing Continuous Integration (CI) pipeline to ensure that new versions of your Workflow are compatible with every change pushed to your version control management. By replaying existing Workflow Histories, you can validate that they do not break determinism and prevent failures from occurring in production.

To use the Replayer in your CI pipeline, consider the following steps:

  1. Fetch a representative sample of Workflow Histories to replay against your new Workflow version. For example, you could fetch the 100 most recent Histories for your Workflow and replay all of them.
  2. Run the Replayer against each of the fetched Histories to ensure that the new Workflow version is compatible with each of them.
  • If the Replayer finds any incompatible changes, the CI pipeline should fail to prevent these changes from going into production. In this case, you should identify and fix the issues before retrying the CI pipeline.
  • If the Replayer does not find any incompatible changes, the CI pipeline should succeed. You can then proceed with confidence that your changes are compatible with your existing Workflow Histories.

What happens if I don’t use the Workflow Replayer?

If you don’t use the Workflow Replayer and your Workflow contains non-deterministic code, you may see a non-determinism error.

This may mean that your new code is incompatible with existing Workflow Histories.

If you are making incompatible code changes to your Workflows and have to deal with Workflow versioning, see Workflow Versioning Strategies:

In the previous sections, you learned the mechanics of how to effectively employ the Workflow Replayer, integrate it within your Continuous Integration (CI) systems, and the potential ramifications of bypassing this crucial step.

Understanding and implementing the Replayer can be a pivotal step in enhancing the reliability and stability of your deployments.

--

--

Patrick Rachford

Currently at Temporal, previously at AWS. I enjoy documenting my thoughts and sharing ideas with others.