Job queues (track B2)

job_queue_jobs

gauge

framework, queue, state

Jobs per state. Standard states: waiting (ready to run), active (held by a worker), scheduled (run later), retry (failed, will run again), dead (retries exhausted, parked). Framework extras: blocked, pending, cancelled. Completed jobs are never counted.

job_queue_oldest_waiting_age_seconds

gauge

framework, queue

Seconds since the oldest waiting job became ready; 0 when nothing waits. Only where the store records the time.

job_queue_stalled_jobs

gauge

framework, queue

Jobs held by a worker past the framework's lease (Laravel reserved, RQ started registry, Asynq lease, Celery visibility) or, without a lease, running longer than stall_after (default 1h).

job_queue_workers

gauge

framework, queue

Live worker processes; queue="*" when the store does not split them per queue.

job_queue_processed_total, job_queue_failed_total

counter

framework, queue

Only where the framework keeps counters.

wt.integration.up

gauge

integration

1 per healthy collection round, 0 when the store cannot be read.

pgboss

Postgres

<schema>.job per name and state (v9 and v10+ columns); age from start_after

dsn, schema (default pgboss)

graphile_worker

Postgres

<schema>.jobs per task identifier: locked (active), attempts exhausted (dead), future run_at (retry or scheduled); lock older than stall_after is stalled

dsn, schema (default graphile_worker), stall_after

oban

Postgres

<prefix>.oban_jobs per queue and state (available, executing, scheduled, retryable, discarded, cancelled); executing longer than stall_after is stalled

dsn, schema (Oban prefix, default public), stall_after

river

Postgres

<schema>.river_job per queue and state (available, running, scheduled, retryable, discarded, pending, cancelled); running longer than stall_after is stalled

dsn, schema (default public), stall_after

good_job

Postgres

<schema>.good_jobs unfinished jobs by queue: performed (active), errored and rescheduled (retry), future (scheduled), ready (waiting); finished with an error (dead); performed longer than stall_after is stalled

dsn, schema (default public), stall_after

solid_queue

Postgres

solid_queue_ready/scheduled/blocked/claimed/failed_executions per queue; claimed by a dead or silent (5 min) process is stalled; live Worker processes

dsn, schema (default public)

que

Postgres

<schema>.que_jobs (Que 1.x and 2.x) unfinished per queue; active = advisory-locked by a worker; expired (dead); que_lockers.worker_count

dsn, schema (default public)

sidekiq

Redis

queues, queue:<q> (depth, oldest enqueued_at), schedule/retry/dead (queue="*"), live processes and their busy, stat:processed/stat:failed

addr, password, username, tls, db, prefix (redis-namespace, e.g. myapp:)

resque

Redis

<prefix>queues, queue:<q> depth (no enqueue time in payloads, so no age), failed list (dead), workers and busy worker:<id> keys, stat:processed/stat:failed

addr, password, tls, db, prefix (default resque:)

celery

Redis

per configured queue the list and its priority lists; unacked (active) and unacked_index entries older than stall_after (stalled). On RabbitMQ use the rabbitmq module

addr, password, tls, db, queues (required), prefix (kombu global_keyprefix), stall_after (= visibility_timeout)

rq

Redis

rq:queues, rq:queue:<q> (depth, oldest job's enqueued_at), registries rq:wip (active; entries past their expiry are stalled), rq:scheduled, rq:deferred (blocked), rq:failed (dead), rq:workers:<q>

addr, password, tls, db, prefix

dramatiq

Redis

per configured queue <ns>:<q> (depth, oldest message_timestamp), .DQ (scheduled), .XQ (dead), in-flight ids of live workers (active), __heartbeats__ (workers). On RabbitMQ use the rabbitmq module

addr, password, tls, db, queues (required), prefix (default dramatiq:)

horizon

Redis

per configured queue queues:<q> (depth, oldest pushedAt/createdAt), :delayed (scheduled), :reserved (active; past retry_after is stalled); Horizon failed_jobs (dead) and supervisor process counts (workers)

addr, password, tls, db, queues (required), prefix (Laravel redis prefix, default laravel_database_), horizon_prefix (default laravel_horizon:)

asynq

Redis

asynq:queues, per queue asynq:{q}:pending (depth, oldest pending_since), active, scheduled, retry, archived (dead), expired lease entries (stalled), processed/failed totals; live asynq:servers (workers)

addr, password, tls, db, prefix

hangfire

SQL Server

JobQueue per queue (unfetched = waiting with the Enqueued state's age, fetched = active), Job in Scheduled / Failed (queue="*"), Server heartbeats in 5 min (workers), stats:succeeded (processed). Redis storage (Hangfire.Pro) is not read

dsn (sqlserver://), schema (default HangFire)

faktory

Faktory server

INFO over the Faktory protocol (HELLO with the salted password hash): queue sizes, Busy (active), Scheduled, Retries, Dead, Workers, total_processed/total_failures. INFO has no job ages

addr (default port 7419), password, tls

temporal

Temporal HTTP API

DescribeTaskQueue with reportStats per configured task queue, workflow and activity summed: backlog (waiting), backlog age, distinct poller identities (workers). Needs a server that reports task queue stats (1.27+)

url (frontend HTTP port 7243), namespace (default default), token (Temporal Cloud API key), queues (required)