📚 Docs Feedback: What should do when we face high volume notification

[triasbrata]

Duplicates

• I have searched the existing issues

Related page

https://docs.novu.co/community/self-hosting-novu/deploy-with-docker

Kind of issue

Missing information

Issue description

No response

Context 🔦

First of all, thank you to the Novu team for the great work!
We're currently using Novu in a self-hosted setup with Docker Compose, and we're aiming to challenge our current configuration to send notifications to up to 600K targets within minutes.
Do you have any insights on which services we should scale to improve the performance of the notification blast?

[linear]

NV-6246 📚 Docs Feedback: What should do when we face high volume notification

[allandiegoasilva]

When you want to send 600,000 notifications in a few minutes, the main bottlenecks are usually the Novu workers, which are responsible for processing the queue and actually triggering the notifications.

To improve performance in your Docker Compose setup, you can:

---

1 Scale the workers

• The novu-worker service processes queued events.
• If you only have one instance, it will process everything serially.
• If you spin up multiple replicas (2, 4, 8...), they will process in parallel, speeding up the delivery.

---

2 Ensure Redis can handle the load

• Redis is used as the queue backend.
• If it's underpowered or has too little memory, it will become a bottleneck.
• For this volume, use a robust Redis setup or even a managed Redis service for better performance.

---

3 Improve database performance (MongoDB)

• Every notification is also stored in MongoDB.
• If MongoDB is slow (due to lack of indexes or insufficient resources), it will delay the processing.
• Make sure your database is well-indexed and has enough resources (CPU, RAM, storage I/O).

---

4 Parallelize API calls

• If you're triggering notifications via the API, avoid sending everything in a single massive request.
• Batch your requests to improve throughput and stability.

---

5 Test and scale gradually

• Don’t try to blast all 600K notifications at once without measuring.
• Scale workers gradually (2 → 4 → 8 replicas) and monitor throughput and resource usage.

---

References

self hosting novu

Docker replicas deploy

[scopsy]

Thank you @allandiegoasilva, I think you summed it in a very good way.

[triasbrata]

Thanks @allandiegoasilva, for the explanation. i will try it