Migrate

From SolidQueue

Both are PostgreSQL ActiveJob adapters — swap the config and gain LISTEN/NOTIFY, DLQs, and recycling.

What changes#

SolidQueue and pgbus share an architecture — a supervisor/worker model, FOR UPDATE SKIP LOCKED polling, forked processes. Both are pure ActiveJob adapters, so your jobs work unchanged. pgbus adds LISTEN/NOTIFY for instant wake-up (SolidQueue only polls), dead-letter queues, worker recycling, and an event bus, and stores messages in PGMQ rather than custom tables.

Effort: low.

Swap the gem and adapter#

Gemfile

# Remove
gem "solid_queue"
gem "mission_control-jobs" # if used

# Add
gem "pgbus"

bundle install && rails generate pgbus:install && rails db:migrate

config/application.rb

config.active_job.queue_adapter = :pgbus # was :solid_queue

If you ran SolidQueue in a separate database via config.solid_queue.connects_to, the pgbus equivalent is config.connects_to — see Separate database.

Convert the worker config#

SolidQueue's processes: N forks N identical workers; in pgbus you list the same worker entry N times (one per process), or use the capsule DSL:

config/initializers/pgbus.rb

Pgbus.configure do |c|
  # SolidQueue: queues "critical", threads 5, processes 2
  c.workers = "critical: 5; critical: 5; default, low: 3"
  c.max_jobs_per_worker = 10_000
  c.max_memory_mb       = 512
end

Configuration mapping#

SolidQueue	pgbus	Notes
`polling_interval`	`polling_interval`	Defaults to 0.1s; LISTEN/NOTIFY makes it a fallback.
`threads`	`threads`	Same concept.
`processes: N`	Repeat the worker entry	One entry per forked process.
`queues: "a,b"`	`queues: [a, b]`	Array, not a comma string.
`queues: "*"`	List queues explicitly	PGMQ queues are explicit.

SolidQueue's limits_concurrency has a near-identical pgbus API (auto-included, no explicit require). See Concurrency & uniqueness; recurring.yml maps to Recurring tasks.

What you gain#

LISTEN/NOTIFY — instant wake-up instead of polling latency.
Dead-letter queues — SolidQueue has none; pgbus routes exhausted jobs to a _dlq queue.
Worker recycling — memory, job-count, and lifetime limits.
An event bus — pub/sub with topic routing, on the same infrastructure.