Skip to content

More optimal way to cleanup/delete blob from external storage, with tradeoff #146

Closed

Description

@longquanzheng

Currently the blobs are only deleted after the whole workflow's history are deleted (all runs -- including runs from continueAsNew, reset and IDReusePolicy).

This is the safest way, mainly because Temporal/Dex support "Reset/TimeTravel" API to go back to any previous point of the history.

But this could potentially lead to some extra storage cost that user may want to optimize.

E.g. if user can accept that they won't reset to any point before X days ago, then we can delete the blobs more aggresively -- we could find the closed runs that closed more than X days ago, if any of the blobs are no longer be used in next run, then they can be safely deleted.

Activity

  1. added
    type:enhancementImprovement to existing behavior, performance, or diagnostics
    P3Low priority; do when convenient
    size:LA week or more, or needs design
    needs-decisionNeeds an owner decision before implementation
    on Oct 26, 2026
  2. longquanzheng commented on Nov 9, 2026

    @longquanzheng
    ContributorAuthor

    The Go blob model was removed with the Go Server (#1059). server-zd deletes superseded values and old history by window and purges objects with the run. Closing; open a new issue for a configurable retention if needed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    P3Low priority; do when convenientneeds-decisionNeeds an owner decision before implementationperf-optimizationsize:LA week or more, or needs designtype:enhancementImprovement to existing behavior, performance, or diagnostics

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions