Statefulset vs Deployment

[cdancy]

Hey All,

Is there a really good reason to use the statefulset option (Aggregator) to deploy vector vs the deployment option (Stateless Aggregator)? I assume you all are persisting something, which would make the Aggregator option viable, but what is it exactly your storing and is that necessary, in your opinion, for production deployments?

Thanks,
Chris

[jszwedko]

Hi @cdancy ,

Persistence is used for:

• Disk buffers
• Components that checkpoint (file, kubernetes_logs, and journald)

If you aren't using any of those, or don't care if the state is lost, then using the stateless Aggregator is preferable to simplify scaling.

[cdancy]

@jszwedko our current setup is that we have fluentbit sending logs to vector with vector routing logs to loki/s3/etc/etc. The heavy handling is done on the fluentbit side with vector acting as the "main log router" if you will. In this scenario it feels like we're OK and don't necessarily need a statefulset unless I'm missing something.

[jszwedko]

It doesn't seem like you need it. You might find these docs helpful: https://vector.dev/docs/setup/going-to-prod/