Data Center · Nexus Dashboard
Right-size a Nexus Dashboard cluster for your fabric: node count, flavor (physical vs virtual), storage per node, and which services (Insights, Orchestrator, Fabric Controller) fit on one cluster.
Nexus Dashboard is the Kubernetes overlay that runs ND Insights (analytics), ND Orchestrator (multi-site), and ND Fabric Controller (DCNM replacement). Under-size it and Insights lags behind the fabric. Over-size it and you've wasted a rack.
You get recommended node count, node flavor (physical or VM specs), total storage, and whether your service mix fits. Insights is the storage hog (~1 TB per 10k endpoints). Orchestrator is light. NDFC scales with switch count.
ND Insights streams telemetry (sFlow, NetFlow, syslog, SNMP) from every leaf. Retention is 30-90 days by default, with a 1-year option for compliance. The data plane is Kafka + Druid + Postgres — all stateful, all disk-bound. Small fabric (<20 leaves): std 3-node works. Mid fabric (20-100 leaves): std 3-node large VMs. Enterprise fabric (>100 leaves): large 6-node physical.
Orchestrator is the light service — mostly templating. NDFC needs bandwidth to switches (POAP, image distribution). Running all three on one cluster is supported up to ~100 switches.
Yes — ND supports cluster expansion (3→6 nodes) in place. But service migration between cluster types is a reinstall.
ND supports cluster backup to external storage. For cross-site HA, deploy a second cluster and pair at the Orchestrator level.
Recommended: 3-node physical cluster (ND-PHYS-SVR-G5 equivalents, each 3×3.2 TB NVMe, 256 GB RAM, dual 25G). Storage ~30 TB total (includes Insights retention). Orchestrator fits alongside.
Recommended: 3-node physical. NDFC alone scales well but you'll want large nodes (384 GB RAM) for image distribution throughput.