"n8n may have run out of memory" issues that started this week

Questions
1

Describe the problem/error/question

Starting this week, we’ve been experiencing a critical memory issue. We have an automation that has been running once per day for around 3 months now that loops through contacts in our CRM and dispatches an AI agent call through Telnyx for follow-up calls. It has been running fine up until this week, when we started getting the error: “Execution stopped at this node - n8n may have run out of memory while running this execution.”

Worse, when trying to view the execution history, the execution data is not visible or saved. The nodes that already ran just say: “Can’t show data. The execution was interrupted, so the data was not saved. Try fixing the workflow and re-executing.”

image
image2517×1226 90.4 KB

Because of this, we can’t do any manual data recovery or continuation of the workflow. It has been causing significant issues this week, and we’ve been trying to reconcile our records and catch up on any call workflows that might have errored midway through.

Last Wednesday and yesterday, it ran without any issues. But when it does, it stops and the error occurs on a random node. There has been no changes to our workflow or n8n instance since this automation started, aside from the frequent n8n instance updates. There are no other automations running on our instance.

Was there a recent change in the n8n cloud subscription limits or memory usage/allocation? This has been quite the challenge this week.

Thanks.

What is the error message (if any)?.

image
image2511×1217 157 KB

Execution stopped at this node
n8n may have run out of memory while running this execution. More context and tips on how to avoid this

Please share your workflow

This project involves multiple workflows and subworkflows. While we want to provide as much context as possible, we cannot share the files publicly because they contain our internal company connections and setup details. Happy to provide these through a support email if possible.

Here is a sample of a workflow that errored out.

image
image1920×1020 293 KB

Share the output returned by the last node

The last node has no output and only shows the “Execution stopped at this node” error banner.

Information on your n8n setup

  • n8n version: 2.21.3
  • Database (default: SQLite): sqlite
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app): n8n Cloud
  • Operating system: Windows 11 Pro
3

Can you revert to previous working n8n version?

4

Hi, I don’t think I can downgrade the version through the n8n dashboard - only toggle between the latest beta or stable.

image
image1235×444 29.4 KB

5

I am facing the same issue after updating N8N this week. The same “ran out of memory” error is occurring. The workflows had been working fine for the past 6 months, but now all workflows are being deactivated because of this issue.

I also reduced the load, but the same problem is still happening. Previously, I was handling much heavier workloads, and N8N never crashed. Now, even while processing only around 60% of that previous load, it is still crashing.

6

Hi, I know this problem – classic memory limit from uncontrolled item load on n8n Cloud.

The issue is in how your workflow processes the CRM contacts. It’s solvable, but the right fix strategy depends on how your loop is currently set up and how many contacts are running simultaneously.

Two questions:

  • How many contacts does the workflow process daily approximately?

  • Are you using a Split-in-Batches node or does everything run in a single pass?

Then I can tell you exactly what needs to be changed.

7

The timing (started this week after working fine for months) combined with multiple people seeing it points to a regression in a recent n8n update. Two things worth checking right now:

  1. Check your n8n version against the changelog - if you’re on n8n Cloud it may have auto-updated. Look for any changes related to execution data handling or AI node memory management around the version you’re on now.

  2. For the workflow itself: the AI agent call looping through CRM contacts is the likely memory spike point - each agent turn retains execution context in memory until the run completes. If the batch size isn’t controlled (e.g., all contacts running in parallel via a Loop node without Split in Batches), an n8n update that changed how execution snapshots are stored could cause a sudden spike. Try wrapping the AI agent call in a Split in Batches node with a batch size of 10-20 and see if the memory stabilizes.

8

@glenbenatiro, good morning!
it’s not possible to confirm that it’s a bug, first compare whether all the volumes are the same between what was working before and now (e.g. number of contacts processed, payload size) and I would disable the saving of successful executions in production.