Skip to content

Issue #7992 : Add load-balancing pipeline and workflow engines - #7999

Closed
mattcasters wants to merge 2 commits into
apache:mainfrom
mattcasters:issue-7992
Closed

Issue #7992 : Add load-balancing pipeline and workflow engines#7999
mattcasters wants to merge 2 commits into
apache:mainfrom
mattcasters:issue-7992

Conversation

@mattcasters

Copy link
Copy Markdown
Contributor

Fixes #7992

This is a first version of load-balancing pipeline and workflow run configurations, not a new transform. Each execution is assigned to one Hop Server from a configured group. To fan out batches, use Pipeline Executor or Workflow Executor with this run configuration.

What it does

  • Table of Hop Servers (name, enabled, max concurrent) in the run configuration editor
  • Even-load vs pack (keep as few servers busy as possible) selection
  • Live status probe (load, occupying unfinished pipelines/workflows, graceful shutdown)
  • Server-side max_concurrent admission (SERVER_AT_CAPACITY) so the cap is transactional
  • Client retry/wait when the pool is full or a submit is refused
  • Re-reads the run configuration so enable/disable and caps can change without restarting the dispatcher
  • Optional assignment state files plus execution-info details
  • Extends the existing remote engines rather than replacing them

Also in this PR

  • Pipeline Executor and Workflow Executor run-configuration widgets are now MetaSelectionLine
  • Get Server Status can output finished pipeline/workflow counts
  • Integration test project load-balance with two long-lived Hop Servers: 100 child pipelines and 100 child workflows, then a post-run status check

Docs

  • xref pages for the native load-balancing pipeline and workflow engines
  • REST max_concurrent on register pipeline/workflow

Introduce Hop Server pool run configurations that assign each pipeline
or workflow execution to one server from a group (even-load or pack),
with live status probing, server-side max_concurrent admission, and
client retry when a server is at capacity.

Also switch Pipeline/Workflow Executor run-configuration widgets to
MetaSelectionLine, add finished execution counts to Get Server Status,
and add a load-balance integration-test project.
…hree tabs

Split the engine-specific widgets into Management, Load-balancing, and
Hop Servers so the metadata editor no longer overflows a single form.
@hansva

hansva commented Aug 19, 2026

Copy link
Copy Markdown
Contributor

This was added together with #8018

@hansva hansva closed this Aug 19, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Feature Request]: Enable distributed batch processing using a pool of Hop Servers

2 participants