Help for owners
n8n is slow or backing up and I am told I need queue mode and workers
Short answer
In queue mode the main n8n instance receives triggers and webhooks and hands each execution to Redis. Separate worker processes run them. It needs a database every process can reach, PostgreSQL is recommended, and the same encryption key on every worker. Without those, workers start but cannot run your workflows.
Does this sound like your situation?
- n8n gets slow or unresponsive when several workflows run at once.
- Webhooks time out while a heavy workflow is running.
- My developer says I should add workers and Redis.
- I set up workers and nothing runs, or the credentials are rejected.
- I do not know if I need queue mode at all.
Why does this happen?
One process does everything
In the normal mode a single n8n process takes the triggers and runs the executions. When one slow or heavy workflow takes the capacity, everything else waits, including webhook calls.
Queue mode splits the job into three parts
The main instance handles timers and webhook calls and creates an execution without running it. It passes the execution ID to Redis, which keeps the queue of pending executions. Workers pick executions from Redis, load the workflow from the database, run it and write the result back.
Confirmed by docs.n8n.io
SQLite is not a fit for it
n8n states that running queue mode with an SQLite database is not recommended, and points to its supported PostgreSQL versions. Every worker needs to reach the same database as the main instance.
Confirmed by docs.n8n.io
Workers cannot read credentials without the same key
The encryption key of the main instance has to be given to every worker and webhook processor. Without it they cannot read the credentials stored in the database, so workflows fail at the first step that needs a login.
Confirmed by docs.n8n.io
Files and concurrency need a plan
In queue mode binary data is stored in the database by default. Filesystem storage needs the same shared volume at the same path for every process. Each worker runs several jobs at once, and n8n warns that low concurrency across many workers can exhaust the database connection pool.
Confirmed by docs.n8n.io
What can I do about it today?
- Describe the slowdown in one sentenceFor example: webhooks time out at 9 a.m. when the import runs. A concrete picture tells a developer whether the problem is one heavy workflow or general load.
- Find out which database you useAsk whether n8n runs on SQLite or PostgreSQL. This decides how big the change is.
- Do not copy a queue-mode guide into your live serverA half-finished change can stop every automation. Ask for a rehearsal on a copy first.
- Write down the workflows that cannot miss a runPayments, orders and customer messages come first. They define what must be tested after any change.
- Make sure the encryption key is stored safelyAsk where it is kept. Without it, credentials cannot be restored after a failure.
When should I bring someone in?
- Webhooks from customers or payment providers are being missed.
- You are on SQLite and a scaling change has been suggested.
- Workers were added and workflows now fail, or the instance would not start again.
How do I handle it?
- Find the real bottleneckI look at which workflows are heavy and when, so that I know whether one workflow needs restructuring or the instance needs workers.
- Back up firstData folder, database and the encryption key are saved before the mode changes.
- Rehearse on a copyI build main, Redis, database and one worker on a copy and run your most important workflows through it.
- Switch with a way backThe production change is made in a quiet window with the old setup kept, so it can be reverted.
- Set limits and watch themI choose a worker count and concurrency that the database can carry, and add health checks so a dead worker is noticed.
The price depends on what I find, so I quote in writing after a short discovery call. Tell me what is happening and I reply within 24 hours.
What do owners ask about this?
Do I always need queue mode?
No. It exists for scaling. If one process copes with your load, the extra pieces only add things that can fail.
Can I run it on SQLite?
n8n says running queue mode with SQLite is not recommended.
Do webhooks need anything special?
n8n describes optional webhook processors with a load balancer in front, sending its webhook paths to them and everything else to the main process.
What happens to a worker that stops?
The queue lives in Redis and the results in the database, so health checks and restarts matter. This is part of what I set up.
Sources
How was this page made?
I drafted this page with AI assistance from the official documents listed under Sources. Found a mistake? Tell me.