Production Checklist

Infrastructure

• Linux kernel 5.11+ (full io_uring support)
• NVMe storage for Data Plane I/O
• Sufficient locked memory (ulimit -l unlimited)
• Memory limit set appropriately (NODEDB_MEMORY_LIMIT)
• Data Plane cores configured (NODEDB_DATA_PLANE_CORES)

Security

• TLS enabled for all protocols
• Authentication configured (SCRAM, JWT, API keys, or mTLS)
• RBAC roles defined and assigned
• RLS policies for multi-tenant data
• Audit logging level set (at least standard)
• Default passwords changed

Replication

• Replication factor >= 3 for production data
• Cluster has odd number of nodes (3, 5, 7) for Raft quorum
• Cross-region learner replicas if needed

Monitoring

• Prometheus scraping /metrics
• Grafana dashboards configured
• Health check endpoint monitored (/health/ready)
• Alerting on key metrics (WAL fsync latency, replication lag, memory pressure)

Backup

• Regular tenant backups scheduled
• Backup validation (DRY RUN) tested
• Restore procedure documented and tested
• WAL archiving configured for PITR

Operations

• Rolling upgrade procedure documented
• Shard rebalancing tested
• Failure recovery tested (single node, minority failure)