CompleteJobsFromWaitGroup

Marks a batch of jobs as completed. Each entry in jobs has a job_id and an optional metadata map. A job_id is recorded once; reporting the same id again is a no-op for completed_jobs (the call is idempotent per job_id, and the metadata stored is the one from the first time the id was reported).

When completed_jobs reaches counter the group becomes completed (a terminal status) and is scheduled for deletion after its delete_after_finished_seconds retention window.

job_id values are unique within a single wait group. They are independent of process_id used by lock/semaphore leases — see Wait Groups. The optional per-job metadata is opaque to Grackle and is returned by ListWaitGroupCompletedJobs — see Metadata.

Request

{
  "namespace_name": "pipelines",
  "wait_group_name": "batch_2026_06_12",
  "jobs": [
    { "job_id": "shard-0", "metadata": { "worker": "worker-7" } },
    { "job_id": "shard-1" },
    { "job_id": "shard-2" }
  ]
}

Response

• Returns NotFound if the namespace does not exist.
• Returns NotFound if the wait group does not exist.
• Returns InvalidArgument if the call would push completed_jobs above counter.

{
  "wait_group": {
    "name": "batch_2026_06_12",
    "status": "ACTIVE",
    "counter": 110,
    "completed_jobs": 3,
    "version": 1,
    "created_at": 1718150400000000000,
    "updated_at": 1718150480000000000,
    "expires_at": 1718236800000000000,
    "delete_after_finished_seconds": 3600,
    "finished_at": 0,
    "last_activity_at": 1718150480000000000
  }
}