Skip to main content

Evolving Workflows

Workflow evolution is crucial for long-lived systems, especially when workflows are expected to run continuously over extended periods. Workflows4s aims to handle workflow evolution through a combination of strategies:

  1. Checkpointing: Marking segments of workflows as checkpointed, preventing the need for replaying those parts and enabling easier modifications.
  2. Journal Rewrites: Modifying the persisted event journal.
  3. Versioning: Running different workflow versions simultaneously.

Compatibility Guide​

Use the decision flowchart below to determine the optimal strategy for evolving your workflows:

Impact of Workflow Changes​

Below is a detailed explanation of how workflow modifications impact historical (already executed) workflows:

  1. Adding a Step:

    • The newly added step will execute upon workflow wakeup, and its event will be appended to the end of the journal.
    • Currently, this might lead to incorrect recovery, as the event will be appended to the end of the journal. Future enhancements to persistence/recovery mechanisms might address this issue.

  2. Removing a Step:

    • Events previously recorded for the removed step will be ignored during recovery.
    • Workflow definition is always subject to typesafety constraints, hence next step input and logic has to accomodate the potentially missing data.

  3. Modifying Events:

    • Changes to events, either in persisted or runtime form, typically don't affect workflows as long as events remain readable from the journal.

  4. Modifying Behavior:

    • Functions within workflows (logic execution, state rebuilding, etc.) can safely be altered without backward compatibility concerns, provided the modifications remain logically consistent.

Checkpointing and Recovery​

For detailed instructions on checkpointing mechanisms and recovery procedures, visit the Checkpointing documentation.

Journal Rewrites​

Journal rewrites involve directly altering the persisted event log to achieve a specific evolution outcome. Such operations include modifying, adding, or removing events. Currently, Workflows4s does not directly support journal rewrites due to their inherently complex and system-specific nature. Users performing journal rewrites should do so with caution and thorough testing.

Automated Compatibility Detection (Planned)​

Workflows4s may introduce tooling for automated detection of breaking changes, inspired by existing solutions such as Buf.

Follow the progress and discussion on the dedicated GitHub issue #48.