Contents
At a glance
- Category
- Automation mechanics
- Purpose
- Decouple work creation from work processing
- Solves
- Bursty volume, rate-limited downstream services, and work that must survive a crash
- Dead-letter queue
- Where items go after retries are exhausted, so failures are visible not lost
- Cheap version
- A table with a status column plus a scheduled worker workflow
- Monitor
- Queue depth and oldest-item age, not just error counts
What a queue does
A queue sits between the part of a system that creates work and the part that does it. Items go in, workers take them out and process them. The value is decoupling: the producer does not have to wait for the consumer, so a burst of two hundred webhooks in a minute becomes two hundred items processed steadily rather than two hundred simultaneous executions fighting over a rate limit.
Most small-business automations start without one and are fine. You need one when three things appear: bursty volume, a downstream service with limits, or work that must not be lost if the processor is briefly down.
The three problems it solves
Burst absorption. A campaign goes out and four hundred replies arrive in ten minutes. Without a queue, four hundred concurrent runs hit an API that permits sixty a minute and most of them fail. With one, they drain at a rate the API accepts.
Durability. If the processor crashes mid-run, a queued item stays queued and is retried; an in-flight execution simply disappears. Ordering and pacing. Some work must happen in sequence, or must be spread out - outbound messaging is the obvious case, where sending five hundred texts in one second is both a rate-limit problem and a good way to look like a spammer to a carrier.
Dead-letter queues
The companion concept, and the one worth adopting even in simple systems. A dead-letter queue holds items that failed repeatedly - after retries are exhausted - so they are set aside for inspection rather than lost or retried forever. It converts a silent failure into a visible list somebody can work through.
In a no-code platform you rarely have a real dead-letter queue, but you can build the equivalent in ten minutes: an error branch that writes the failed item and its error to a table, plus an alert when that table is not empty. That single addition catches the entire class of "we did not find out for a week".
Cost and how to get one
You may already have one. n8n offers a queue mode for scaling, where executions are distributed to worker processes via Redis. Cloud platforms provide managed queues. And for modest volumes, a database table with a status column and a scheduled worker workflow is a perfectly respectable queue that costs nothing extra.
The cost of adding one is complexity: another moving part, another place work can get stuck, and a need to monitor queue depth. Do not add one speculatively. Add it when you can point at the burst that failed, the outage that lost work, or the rate limit you keep hitting.
What you should know
Monitor depth, not just errors. A queue that is growing steadily means consumption is slower than production, and the system will eventually fall over even though nothing is currently erroring. Queue depth and oldest-item age are the two numbers that tell you this early, and they belong in your observability dashboard.
Queued items also need idempotency, because most queues deliver at least once and a worker that crashes after acting but before acknowledging will see the item again.
Queue FAQ
What is a queue in automation?
A holding area for work items between the thing that creates them and the thing that processes them. It lets a sudden burst of events be handled steadily instead of all at once, and it stops work disappearing if the processor is briefly unavailable.
Do I need a queue for my automation?
Not usually at first. You need one when volume is bursty, when a downstream service has rate limits you keep hitting, or when losing work during an outage is unacceptable. Adding one speculatively just adds a moving part to monitor.
What is a dead-letter queue?
Where items go after retries are exhausted, so they are set aside for someone to inspect rather than lost silently or retried forever. It is the difference between finding out about failures immediately and finding out when a customer complains.
How do I add a queue without extra infrastructure?
A database or platform table with a status column, plus a scheduled workflow that picks up pending rows, processes them and marks them done. It is a genuine queue for modest volumes, costs nothing extra, and gives you a dead-letter list for free.
What should I monitor on a queue?
Depth and the age of the oldest item. A queue growing steadily means you are consuming slower than you are producing, which will eventually fail even though nothing is erroring right now. Error counts alone will not show you that.
Do queued items need idempotency?
Yes. Most queues deliver at least once, and a worker that completes an action then crashes before acknowledging will receive the same item again. Without idempotency, that is a duplicate message or record every time a worker dies at the wrong moment.
Primary source
n8n - queue mode documentation
Checked August 2026. Anything priced or rate-limited moves, so verify against the source before you budget.
Want this built rather than explained?
Knowing what queue means is one thing; having it running in your business is another. We scope one workflow at a time, quote a fixed price in writing after a free consult, build it, and hand it over with documentation you own.
