@Rocky What is the minimum recommended cluster size for production, with a shared none arquitecture?

For a production environment using StarRocks' shared-nothing architecture, the minimum recommended cluster size is: * FE (Frontend) Nodes: 3 Follower FE nodes. This ensures High Availability (HA) using the Raft protocol. If the Leader FE fails, a new one can be elected as long as a majority (2 out of 3) are alive. * BE (Backend) Nodes: 3 BE nodes. By default, StarRocks maintains 3 replicas for data tablets to ensure data reliability and service availability. A 3-BE setup allows the cluster to remain functional even if one node fails. Recommended Hardware (Production) For these nodes, the minimum recommended resource allocation per node is: * FE: 8 CPU cores, 16 GB RAM, 100 GB+ SSD/HDD. * BE: 16 CPU cores, 64 GB RAM, High-performance SSD (recommended for storage). Key Notes: * Avoid SPOFs: While StarRocks can run on a single node (1 FE + 1 BE) for testing or development, it is not recommended for production because it provides no redundancy or failover capability. * Colocation: In small production clusters, you can co-locate one FE and one BE on the same physical/virtual machine (total of 3 machines), but for high-load environments, it is better to deploy them on separate instances to avoid resource contention. For more details, you can refer to the official StarRocks Planning a Cluster documentation. References • deployment/plan_cluster.md • faq/Deploy_faq.md • knowledge/usage/best-practice.md • faq/Others.md • introduction/Architecture.md

is very huge, give me a smaller cluster

You need to keep tagging it for follow-up questions. If you're not doing high volume partial updates, shared-data might actually be better for you. Gives you more elasticity and free durability via S3