Harvester Live Migration: Disk I/O Consistency During VM Transfer

Goal

Move a small, write‑heavy VM from one Harvester host to another without service interruption.

Unresolved Constraint

The documentation confirms that live migration streams disk I/O to the target, but it does not specify the consistency guarantees for write‑heavy workloads. In practice, users have reported transient I/O stalls and occasional data duplication when migrating VMs that run database services.

Key Questions

• What consistency guarantees does Harvester provide for disk I/O during live migration of VMs running write‑intensive workloads?
• Does enabling the blockIO migration option effectively mitigate transient I/O stalls, and what latency threshold is considered acceptable?
• Are there any rollback or pause/resume semantics available for the migration process to prevent data duplication or corruption?