hydrakit / queue

v0.31.2 2026-10-07 19:47 UTC

Requires

• php: >=8.4
• hydrakit/console: ^0.31
• hydrakit/core: ^0.31
• hydrakit/database: ^0.31
• hydrakit/scheduler: ^0.31
• psr/clock: ^1.0
• psr/container: ^2.0
• psr/event-dispatcher: ^1.0
• psr/log: ^3.0

Requires (Dev)

• ext-pdo_sqlite: *
• hydrakit/event: ^0.31
• hydrakit/log: ^0.31
• phpunit/phpunit: ^11.0

MIT

• William Hleucka

Published on Packagist.org on 2026-10-09 13:58 UTC

README

Part of the Hydra PHP framework. Documentation: hydra.williamhleucka.com/docs.

Read-only mirror. hydrakit/queue is developed in hydra-foundation/hydra under packages/queue, and republished here on every push. A commit pushed to this repository is overwritten by the next one; issues are disabled for that reason, and a pull request opened here cannot be merged. Both belong upstream.

Work that outlives the request. A JobInterface is built by the container and handed the payload it was pushed with; QueueInterface::push() takes the job's class, a payload of scalars, nulls and arrays, stored as JSON, and an optional delay in seconds.

DatabaseQueue keeps jobs in a jobs table and failures in failed_jobs. There is no daemon: Worker is a scheduler batch, $schedule->drain(Worker::class)->everyMinute(), that claims up to twenty jobs at a time. A job that finishes is deleted. One that throws is tried again after 10 seconds, then 60, and after its third failure moves to failed_jobs with the exception. A claim whose worker died is taken again after fifteen minutes, and the attempt is counted when the job is claimed, so a job that kills its worker still runs out of tries.

queue:failed lists what ran out of tries, newest first, with the exception's message. queue:retry <id> puts one back on the queue, due now and with its tries restored; queue:retry --all puts back every one. queue:forget <id> deletes one without running it, and queue:flush deletes them all.

Testing\FakeQueue records pushes for assertQueued(), and refuses whatever the database queue would.