Hey everyone — I’m fairly new to n8n, but I have a lot of experience building marketing automations in tools like Clay, Zapier, and Salesforce.
I built a workflow that uses OpenAI, a few custom HTTP requests, and Salesforce connections. I noticed that when an execution fails, the workflow appears to keep re-running automatically. In this case, the failure was caused by hitting the OpenAI token limit, and it ended up re-running ~10 times overnight.
I’ve double-checked that none of the nodes are configured to “Retry on fail,” and the Zapier webhook that triggers the workflow did not re-fire (I included screenshots of both the n8n execution history and the Zapier trigger history showing no re-triggers). The repeated runs also cost us money, since it was maxing out token usage over and over again.
Has anyone run into this behavior before, or have ideas on what could cause repeated executions after a failure? Any advice would be much appreciated.
Thank you for the screenshot, it means this workflow is triggered from Zapier via HTTP call.
I’m not familiar with Zapier but there might be a possibility that the Zapier is the one that does the retry automatically. So try to check that.
What is the value of “Respond” from the “Webhook” node being set to? (should be either “Immediately” or “When Last Node Finishes”. If “When Last Node Finishes”, it might return error code that might be detected as retry-able by the caller.
Additionally, you might want to try to debug the past failed executions and check if the request parameter is similar from the caller or further try to add a timestamp to the request body for debugging purpose to see if it’s the caller (Zapier) or n8n worker issue.
