Memory usage keeps growing until process crash

Hi all,
I’m having a strange issue with version 0.198.2 of n8n, that I wasn’t experiencing with the older versions. I am running n8n using docker-compose and the issue occurs both on mac and linux systems.

I have a workflow which splits 1000 items in batches of 1 item each and processes them. During the process there is an HTTP request to an external service. There is a 5 sec pause between the loops.

The issue is that at each iteration the memory usage keeps growing, until the whole process breaks with the message:

Workflow execution process did crash for an unknown reason!

I am only able to process around 70 items before it crashes.
It’s crashing during the execution of the HTTP request node, because the interface remains stuck on that node.

As I said, the workflow was always completed correctly on previous versions (I don’t remember the exact number, but it was launched using docker about 8 months ago).

I tried the workarounds suggested in this topic but without success:

- N8N_USE_DEPRECATED_REQUEST_LIB=true
- EXECUTIONS_PROCESS=main

Any suggestions?
Thanks for your help!

Hi @4ng3l0, I am sorry to hear you’re having trouble. Unfortunately, it seems you have deleted the forum template for new questions rather than filling it out.

Can you share your workflow as well as example data structures for each node involved? Also, can you confirm from which version you upgraded and confirm the environment variables you have set (typically in your docker compose file)?

The amount of data processed by n8n in newer versions has increased slightly (due to a new feature called paired items) but I am very surprised to hear that instead of 1,000 items you can only process 70 items following your upgrade.