3 months ago
- Workers still OOM-killed – Worker (pid:99) was sent SIGKILL! Perhaps out of memory?
- ERR_CONNECTION_RESET – Connections drop when workers are killed.
Most reliable fix: increase memory to 1GB
512MB is too small for PyTorch + any embedding model. Even MiniLM (~90MB) plus runtime can exceed 512MB.
Steps:
- Railway Dashboard → validify_match service → Settings → Resources
- Set Memory to 1 GB (or 1024 MB)
- Redeploy the service
That should stop the OOM kills and allow matching to run.
If you must stay at 512MB:
- Add EMBEDDING_MODEL=minilm on the service Variables tab (not only Shared Variables)
- Confirm the latest commit is deployed (check the deployment commit hash)
- Redeploy after changing variables
Increasing memory to 1GB is the most straightforward way to resolve this.
Pinned Solution
3 months ago
Add this environment variable to your n8n worker service:
NODE_OPTIONS=--max_old_space_size=4096
This 4096 value represents the maximum amount of RAM it can use, which equals 4GB. Edit this value accordingly (1024 = 1GB of RAM).
Make sure your worker service also has that amount of RAM available by going into its service settings and scrolling down to the "Scale" section.
3 Replies
3 months ago
Hey Dale! It looks like your message might have been generated by an AI assistant. If you have a specific question about your Railway service, feel free to post it in the Help Station as a bounty and the community can help out.
Status changed to Awaiting User Response Railway • 3 months ago
Status changed to Solved sam-a • 3 months ago
sam-a
Hey Dale! It looks like your message might have been generated by an AI assistant. If you have a specific question about your Railway service, feel free to post it in the Help Station as a bounty and the community can help out.
3 months ago
Hi Sam. This was the review of the logs. deployment was successful when features failed when used in production.
I am seeking help in terms of optimisation or config
Thanks
Dale
Status changed to Awaiting Railway Response Railway • 3 months ago
3 months ago
Add this environment variable to your n8n worker service:
NODE_OPTIONS=--max_old_space_size=4096
This 4096 value represents the maximum amount of RAM it can use, which equals 4GB. Edit this value accordingly (1024 = 1GB of RAM).
Make sure your worker service also has that amount of RAM available by going into its service settings and scrolling down to the "Scale" section.
Status changed to Solved medim • about 1 month ago