How to set concurrency limits in n8n?

The short, direct answer is: n8n doesn’t let you set per‑workflow concurrency limits inside the workflow itself, but you can control concurrency using three real, production‑safe mechanisms: global execution limits, queue mode + concurrency settings in Redis workers, and Webhook/Trigger rate limiting at the infrastructure level. In real projects, proper concurrency control usually means running n8n in queue mode and tuning concurrency per worker.

 

How concurrency really works in n8n

 

n8n runs workflows as “executions”. When multiple workflows get triggered at the same time (for example via webhook triggers, schedules, or incoming events), n8n may run several executions in parallel. How many can run at once depends on how your environment is configured.

There is no “per workflow concurrency” setting inside the UI, so the way we control concurrency is at the platform level (n8n server settings) or by shaping traffic before it reaches n8n.

 

Real production ways to limit concurrency in n8n

 

1. Global execution limits (simple setups)

• You can set how many workflows n8n is allowed to execute in parallel across the whole instance.
• This works only in “regular mode” (not queue mode).
• You use the environment variable N8N_CONCURRENCY_PRODUCTION\_LIMIT.

export N8N_CONCURRENCY_PRODUCTION_LIMIT=5   // allow max 5 executions at the same time

Useful on small servers where you want to make sure a spike of incoming triggers doesn’t overload the instance. It applies globally — all workflows share this limit.

 

2. Queue mode + worker concurrency (the real production solution)

• You run n8n with separate “main” and “worker” processes using Redis.
• The workers actually execute workflows.
• You can control how many workflows each worker is allowed to run in parallel.

export N8N_EXECUTIONS_MODE=queue
export QUEUE_WORKER_CONCURRENCY=3   // each worker runs max 3 executions at once

• If you need finer control, you run multiple workers with different concurrency settings.
• If you want to slow down a specific workload, you dedicate a low‑concurrency worker and route certain workflows there using Workflow Tags → Worker assignment (a real n8n feature).

This is the most flexible and scalable way to manage concurrency in production deployments.

 

3. Rate limiting before the workflow reaches n8n (infrastructure-level)

• For webhook-triggered workflows, you can use your reverse proxy (NGINX, Traefik, Cloudflare, API Gateway) to limit request rates.
• This effectively controls how many executions hit your n8n instance at once.
• Useful for API-heavy environments or when external systems may flood your endpoint.

# Example NGINX rate limit
limit_req_zone $binary_remote_addr zone=slowzone:10m rate=5r/s;

location /webhook/ {
    limit_req zone=slowzone;
    proxy_pass http://n8n:5678;
}

This protects n8n from traffic spikes and prevents workflow execution pileups.

 

Why you can’t set concurrency inside the workflow itself

 

n8n workflows are stateless execution graphs. Each run is independent. There is no built‑in mechanism that says “only allow 1 execution of this workflow at a time”. The engine doesn’t track a “lock” per workflow.

To mimic per-workflow concurrency, you use either:

• queue mode with a dedicated low‑concurrency worker
• infrastructure rate limits for specific webhook paths

Both are reliable and proven in production environments.

 

What I normally do in real production builds

 

• For high-traffic workflows: queue mode + multiple workers, control concurrency using QUEUE_WORKER_CONCURRENCY.
• For sensitive workflows (like sending emails or updating CRM records): put them in their own worker group with concurrency=1.
• For public webhooks: apply rate limits at the proxy to prevent bursts that overload the instance.

This pattern gives predictable load, prevents timeouts, and avoids broken API quotas.

 

Bottom line: you set concurrency limits in n8n by configuring global execution limits, or the worker concurrency when running in queue mode, or rate limiting traffic before it reaches n8n. n8n does not provide a per‑workflow concurrency setting inside the workflow itself.