Hi team, I need some guidance on Pinot performance related to volume and storage. I'm noticing that as data volume increases, Pinot’s query performance and realtime consumption seem to degrade. I’m also seeing segments frequently going into BAD / UPDATE state. When I pause ingestion, after ~15–20 minutes, the segments slowly move back to a normal state. Here’s my current setup: • StorageClass: expandable-storage • Deep Store: S3 • Server Replicas: 2 • Instance Type:

r5a.xlarge

• Server CPU Request: 250m • Server Memory Request: 6Gi • Server JVM: • • -Xms8G • -Xmx16G • -XX:+UseG1GC •

-XX:MaxGCPauseMillis=200

• Replication Factor: 2 for table Question: How does increasing data volume impact Pinot’s query latency and realtime segment consumption under this setup? And why might segments consistently fall into BAD/UPDATE state unless ingestion is paused? Any insights or tuning recommendations would be very helpful. Server Placement: each server runs on a dedicated machine using node affinity (no other workloads on these nodes)

250m cpu? That is way too low for any distributed system. You can google capacity planning for Pinot, there is a couple of good blogs

250m CPU is just a request. we don't set any limit for cpu and memory as its deployed in separate machine. I beleive we will not have a issue based on it. If still it causes issue ill go head and change it

For segments going into bad state you can check the debug api or server logs.