SQL to create the schema for Arbiter tables

Default PostgreSQL schema name for Arbiter tables

Using a dedicated schema prevents namespace pollution in the user's public schema.

SQL to create the main job queue table within a schema

This table stores pending and in-progress jobs. Completed jobs are deleted, failed jobs are moved to the DLQ.

SQL to create the dead letter queue table within a schema

Jobs that fail repeatedly (exceed max attempts) are moved here for inspection. This table contains ALL the Job fields (complete snapshot) plus DLQ-specific metadata.

SQL to create a partial index on group_key for efficient per-group lookups.

Used by claim queries' LATERAL subqueries and by the DELETE/UPDATE triggers when recomputing in_flight_until. Composite (group_key, priority, id) so the DELETE trigger can recompute min values via index-only lookup.

SQL to create index on DLQ group_key for querying failed jobs by group

SQL to create index on DLQ failed_at for time-based queries

SQL to create index on DLQ parent_id for efficient child lookups (DLQ child counts)

SQL to create unique index on dedup_key for job deduplication

SQL to create a partial index on parent_id for efficient child lookups.

SQL to create the NOTIFY function for a specific table

This function is triggered after job insertion and sends a notification on a table-specific channel. Each table gets its own function and channel for isolation.

Note: The channel name is quoted as a string literal (single quotes), not an identifier.

SQL to create the NOTIFY trigger for a specific table

This trigger fires AFTER INSERT on the job queue table and calls the table-specific notify function.

SQL to drop the NOTIFY trigger

SQL to drop the NOTIFY function for a specific table

SQL to create the event streaming notification function

This is a shared function (one per schema) that fires on INSERT, UPDATE, DELETE of any job table and INSERT on any DLQ table. It sends a lightweight JSON event via pg_notify on the arbiter_job_events channel with just the event type, table name, and job ID.

The function uses TG_TABLE_NAME and TG_OP to determine context, so it works for all tables without per-table copies.

SQL to create event streaming triggers for a table and its DLQ

Creates a combined INSERTUPDATEDELETE trigger on the main table and an INSERT trigger on the DLQ table, both calling the shared notify_job_event function. Also drops any legacy per-operation triggers left by older versions of setupEventTriggers.

Generate notification channel name for a table

Each table gets its own NOTIFY channel for job insertions. Example: notificationChannelForTable "email_jobs" -> "email_jobs_created"

Channel name used by the event streaming (SSE) system.

Per-table NOTIFY trigger function name.

Per-table NOTIFY trigger name.

Shared event streaming trigger function name (one per schema).

Per-table event streaming trigger name.

Per-table DLQ event streaming trigger name.

Qualified table name: jobQueueTable "arbiter" "email_jobs" -> "arbiter"."email_jobs"

Qualified DLQ table name: jobQueueDLQTable "arbiter" "email_jobs" -> "arbiter"."email_jobs_dlq"

Qualified results table name: jobQueueResultsTable "arbiter" "email_jobs" -> "arbiter"."email_jobs_results"

Qualified groups table name: jobQueueGroupsTable "arbiter" "email_jobs" -> "arbiter"."email_jobs_groups"

Qualified reaper sequence name.

SQL to create the reaper coordination sequence. Stores the epoch (seconds) of the last reaper run.

SQL to create the results table for storing child job results.

Child results are stored as individual rows (one per child), keyed by (parent_id, child_id). The ON DELETE CASCADE foreign key ensures cleanup when the parent is acked (deleted).

SQL to create the groups table for fast grouped claims.

Stores one row per distinct group_key with pre-computed min_priority, min_id, job_count, and in_flight_until. Maintained by statement-level AFTER triggers on the main job table (see createGroupsTriggerFunctionsSQL).

SQL to create a ranking index on the groups table.

Non-partial so that UPDATE SET job_count, in_flight_until etc. never change indexed columns — enabling HOT updates on every ack and claim.

Three trigger functions maintaining the groups table via statement-level AFTER triggers. Uses incremental operations where possible.

SQL to create 3 statement-level AFTER triggers on the main job table that call the groups maintenance functions.

Uses REFERENCING NEW/OLD TABLE AS for efficient batch access to affected rows via transition tables.

Quote a PostgreSQL identifier (schema name, table name, column name).

This escapes double quotes by doubling them and wraps the identifier in quotes. This prevents SQL injection when using dynamic identifiers.