Constraints for queue operations (requires table name lookup from registry).

Constraints for job operations (table name stored in job).

The table name payload's entry in this monad's registry declares.

The registry declares both admission policy kinds.

Insert a job. Returns the inserted job, or Nothing if skipped by dedup (IgnoreDuplicate).

Insert multiple jobs in one round-trip. Returns the jobs that were inserted. Dedup'd jobs are excluded. Use insertJobTree for parent-child relationships.

insertJobsBatch returning the count of inserted rows.

Claim visible jobs, at most one per group, and fewer than the limit once the groups run out. Stamps the anonymous claimant. A concurrency pool caps this path as it caps any other claim.

claimNextVisibleJobs under a given worker id. The dispatcher claims this way.

claimNextVisibleJobs claiming up to batchSize jobs from each group, still in per-group order.

Assemble a pool's claim statements once (see mkClaimSql). Batch size 1 is the single-job claim.

Add tokens to a key's bucket, capped at max, and wake any of its jobs parked mid-wait. A no-op without a policy.

Delete reclaimable idle (full) buckets. Returns the number pruned. The worker reaper runs this. A full bucket re-seeds at full on next use.

Reset every bucket for a prefix to full and wake its throttled jobs. Returns the number of buckets reset. A manual (0-refill) policy plus a cron calling this at the boundary is a fixed window.

List every rate-limit policy with its params, bucket stats, and a live count of currently-throttled jobs per prefix across the registry's queues.

One prefix's rate-limit policy with its params, bucket stats, and live throttled count. Nothing when the prefix has no policy.

Whether a rate-limit policy exists for a prefix.

List a prefix's buckets with fill levels, paginated.

List every concurrency pool with its default/override limit and live key and in-flight aggregates.

One prefix's concurrency pool with its default/override limit and live aggregates. Nothing when the prefix has no pool.

List a prefix's keys with effective cap and in-flight fill fraction, paginated.

Set or clear a policy's override params and wake the prefix's parked jobs. Returns rows affected (0 if absent).

Override a policy with this shape, until it is cleared. Returns rows affected.

Drop a policy's overrides. Its declared params apply again. Returns rows affected.

Apply a pool's override-limit patch on its policy row, retuning every key under the prefix live. Lowering it does not preempt in-flight jobs until they drain. Returns rows affected.

Override a declared pool's limit for every key under it. Returns rows affected.

Drop a pool's override. Its declared limit applies again. Returns rows affected.

Delete drained concurrency rows with no live job. The reaper runs this.

Recompute the concurrency counts from live jobs, repairing any trigger drift.

Rebuild the concurrency counts when a crash truncated the UNLOGGED table. The reaper runs this periodically.

Reconcile then prune. The reaper runs this.

Ack a completed job, deleting it or suspending it when its children are still running. Returns 1, or 0 for a job already gone.

ackJob over a batch from one queue in one statement, returning the ids acked. Reclaimed jobs are absent.

Park a failed job for its retry backoff. Returns 0 for a job another worker holds.

Soft-nack a job. It is reprocessed once its visibility timeout lapses. No failure is recorded and no attempt is consumed. Returns 0 for a job another worker holds.

nackJob over a batch from one queue in one statement, returning the ids nacked. Jobs another worker holds are absent.

Extend a job's visibility timeout by hand, for a long-running job. Returns 0 for a job that is gone, reclaimed or suspended. setVisibilityTimeoutBatch tells them apart.

setVisibilityTimeout over a batch from one queue.

Result of setting visibility timeout for a single job in a batch.

A narrowing predicate on a job listing. FilterJobId and the completed_at range name DLQ and archive columns.

List filtered jobs, newest first.

Count filtered jobs.

List filtered DLQ jobs, most recently failed first.

Count filtered DLQ jobs.

Move a job to the DLQ. Returns 0 for a job another worker holds.

Move a batch of jobs to the DLQ, skipping any another worker holds. Returns the number moved.

List DLQ jobs, most recently failed first.

List archived jobs, most recently completed first.

Fetch a single archived job by its original job id.

List archived jobs in a group, most recent first, with pagination.

Delete one archived job by its archive primary key. Returns rows deleted.

Delete archived jobs by archive primary key. Returns rows deleted.

Re-enqueue an archived job as a fresh standalone job, keeping the archive row. Returns the new job, or Nothing if the archive row no longer exists.

Retry a DLQ job, re-inserting it into the queue with a fresh attempt count. Nothing when the DLQ row is gone.

Whether a DLQ job with the given id exists.

Delete a DLQ job. Returns 0 when the row is already gone.

Delete DLQ jobs by id. Returns the number deleted.

List jobs, newest first.

Fetch a job by id.

Whether a job with the given id exists in this payload's queue table.

List a group's jobs.

List a parent's children.

Delete a job by id. Returns 0 for a job with children. cancelJobCascade deletes those.

Delete jobs by id. Returns the number deleted.

Delete a job and its descendants, interrupting any handler running one of them by NOTIFY on the cancel channel.

Make a delayed or retrying job immediately visible. Refuses an in-flight job.

Per-status breakdown of a queue. The per-status counts partition the queue and sum to totalJobs, mirroring the derived job status taxonomy.

A queue's per-status counts and backlog ages.

Count every job in the queue.

Count a group's jobs.

Count a parent's children.

Count the queue's DLQ jobs.

Child counts as (total, paused) per parent id, over a batch. Parents with none are absent.

Count a parent's DLQ'd children.

DLQ child counts per parent id, over a batch. Parents with none are absent.

Suspend a parent's claimable children. In-flight ones are left alone. Returns the number suspended.

Resume a parent's suspended children. Returns the number resumed.

Delete a job and every descendant under it. Returns the number deleted.

Suspend a job, making it unclaimable. Refuses an in-flight job.

Resume a suspended job, making it claimable again.

Insert a child's result, encoded as the queue's declared result type and keyed by (parent_id, child_id). Its foreign key cascades. Acking the parent clears it. Returns 0 for a result that stores nothing.

insertResult with a raw JSON value, bypassing the queue's declared result type. A value childResults cannot decode surfaces there as a Left.

A parent's child results, keyed by child id.

A parent's DLQ'd children's last errors, keyed by child id.

Read all child data for a rollup finalizer. Returns (childId->result, childId->error, parentStateSnapshot, dlqRowId->error).

Merge child results, DLQ errors and the snapshot, left-biased in that order.

Snapshot results into parent_state before DLQ move.

Read raw parent_state snapshot from the DB.

Schema-wide groups-table refresh, correcting every registered queue's summary drift. Walks each groups table to the end, one bounded batch and one transaction at a time. This is a repair operation. The reaper runs refreshAllGroups for a single batch per tick. Returns the rows rewritten and the queue names that failed.

Register or refresh a worker and return its effective pause state.

Record a heartbeat and return the worker's effective pause state.

Set a worker's pause flag.

Mark a worker as gracefully draining.

Remove a worker registry row.

List workers, optionally filtered by queue and heartbeat age.

Delete workers older than their recorded stale threshold.

A row in the worker registry. One row per running worker pool.

Ensure a queue registry row exists.

Set a queue's pause flag and notify its workers.

Get one queue registry row.

List all queues registered in the schema.

A row of the arbiter_queues pause/resume table.

List cron schedules ordered by name, optionally filtered by queue.

Get a single cron schedule by name.

Patch a cron schedule, writing the overrides as given. Returns rows affected, 0 for a name that is not there. Arbiter.Worker.updateCronScheduleChecked rejects overrides the scheduler cannot parse.

Run work at most once per interval across every worker pool sharing the same schema. See runGated.

Insert a JobTree atomically. Returns all inserted jobs (pre-order), or Left if the root has a dedup conflict. Rolls back on any failure.

The schema name for this monad's Arbiter tables.