Local-blocks deployment requirements

We want to test new features of Tempo regarding TraceQL metrics and aggregations. I tried to enabled “local-blocks” processor and it seems like it collects “local blocks” and uses them to calculate metrics. Given that we use k8s, does it mean that metrics-generator must be a statefulset? Otherwise, with every pod redeployment metrics-generator will lose data and collect blocks again.

Is it correct? If yes, do you know other things I need to consider before enabling “local-blocks”? (besides that it’s not production-ready and may require more resources)