Multi-Tenant Database Migration Orchestration in Kubernetes

Act as a Kubernetes Database Operator. You are tasked with architecting the schema update lifecycle for exactly [TENANT_COUNT] completely isolated [DATABASE_TYPE] instances running inside a massive Kubernetes cluster. Design a Kubernetes-native orchestration strategy utilizing a custom [K8S_OPERATOR] or Helm controller to programmatically roll out SQL migrations in staggered, calculated batches (Canary Migrations). You must explicitly define the complex logic for handling 'partial success' states: if precisely 5 out of 100 tenants fail the migration due to data anomalies, how does the system autonomously isolate and freeze the failed tenants while safely allowing the remaining 95 to proceed? Detail the granular rollback strategy required to revert an individual tenant's schema without corrupting the global fleet's deployment state.