Goal

This epic aims to provide administrators with the ability to live migrate specific virtual machines to a designated node. This allows for precise control over VM placement for scenarios like hardware maintenance or performance optimization. The goal is to ensure these targeted migrations are reliable, even in clusters utilizing memory overcommit, and are not dropped due to processing queue limits.

User Stories

• As a cluster administrator, I want to migrate one or more VMs to a specific node, so that I can control VM placement for maintenance or performance tuning.
• As a cluster administrator, I want to migrate VMs to a specific node that is utilizing memory overcommit, so that I can efficiently manage resources in an overcommitted cluster.
• As a cluster administrator, I want to migrate VMs without the migration request failing due to a full processing queue, so that I don't have to manually requeue them.

Non-Requirements

• Testing the functionality of the feature itself is not part of this epic.

Notes

• This feature is intended for exceptional situations and one-off migrations, not as a primary tool for load balancing workloads.
• To protect the system from an unbounded number of migration requests, the migration controller will limit the total number of queued VirtualMachineInstanceMigration (VMIM) objects. The queue size will be limited to double the ParallelMigrationsPerCluster count, and any new requests beyond this limit will fail fast.
• The virtctl migrate command will include a new --addedNodeSelector flag to specify the target node.
• The implementation adds an addedNodeSelector field to the VirtualMachineInstanceMigration spec.