Help for owners
My n8n workflow fails with 'too many requests' when it processes a big list
Short answer
n8n sends one request per input item, so a list of thousands can hit another service's rate limit and return error 429. The fix is to pace the requests: batch them with a pause, use Loop Over Items with a Wait node, or retry. Retry is limited to five tries and five seconds, so long limits need pacing.
Does this sound like your situation?
- The workflow works with ten rows and fails with a thousand.
- The error says the service is receiving too many requests from you.
- It fails partway through and I do not know which rows were done.
- Some customers got a message twice after a rerun.
- The other service sent a warning about my usage.
Why does this happen?
Each item becomes its own request
n8n's documentation says that in the HTTP Request node each input item generates a separate request. A sheet with five thousand rows is five thousand calls, fired as fast as n8n can send them.
Confirmed by docs.n8n.io
The other service sets the limit
An API may limit requests per minute or per day, and the size of what you send or receive in one call. n8n shows the service's own message in the node output, and its guidance is to read that service's documentation for the actual limit.
Confirmed by docs.n8n.io
Retry on Fail is a small tool
n8n says Wait Between Tries is limited to 5000 ms and Max Tries to five, and that the HTTP Request node does not automatically obey the service's Retry-After header. For longer delays it recommends batching or Loop Over Items with a Wait node.
Confirmed by docs.n8n.io
Pacing can still be set wrong
The Batching option has Items per Batch and a Batch Interval in milliseconds. Loop Over Items with a Wait node afterwards, connected back to the loop, splits a list into chunks and pauses between them. A batch size or interval that is too generous will still hit the limit.
Confirmed by docs.n8n.io
A rerun repeats what already happened
When a run stops in the middle, the earlier items were already sent. Running the whole workflow again can send them a second time unless the workflow records what is finished.
What can I do about it today?
- Stop re-running the workflow blindlyIf it sends messages, creates invoices or updates records, a rerun can repeat actions. Check what was already done first.
- Find the other service's limitLook in its API documentation for requests per minute and per day. Write the number down.
- Note how many items the workflow handlesA list size and a limit tell you how long a safe run must take.
- Run it in a quiet periodIf the same service is used by your website or staff, a heavy run can slow them as well.
- Ask for a small test batchRun ten items with the pacing set before the full list.
When should I bring someone in?
- The workflow changes money, stock or customer messages.
- You cannot tell which items were processed.
- The service has warned or blocked your account.
How do I handle it?
- Read the service's limitsFrom its own documentation, including daily as well as per-minute limits.
- Choose a pacing methodBatching in the node, or Loop Over Items with Wait, set safely below the limit.
- Make the run safe to repeatEach item is marked once done, so a stop in the middle never sends anything twice.
- Handle errors per itemA failed item is recorded and the rest continue, with an alert.
- Test small, then fullA trial batch first, then the whole list in a quiet window.
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?
How do I slow n8n down?
Use the HTTP Request node's Batching option with a batch interval, or Loop Over Items with a Wait node.
Does Retry on Fail solve it?
For short limits it can help. n8n caps it at five tries and 5 seconds between them.
Why does n8n not slow down by itself?
n8n's documentation says rate limiting is handled with these settings. The limit is the other service's, so you need to set the pace.
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.