Dell PowerFlex 4.5.x Administration Guide

Migrating vTrees

Migration of a volume tree (vTree) allows you to move a vTree to a different storage pool.

Migration of a vTree frees up capacity in the source storage pool. For example, you can migrate a vTree from an HDD-based storage pool to an SSD-based storage pool, or to a storage pool with different attributes such as thin or thick.

When a user requests a vTree migration, the MDM begins the process by estimating whether the destination storage pool has enough capacity for a successful migration. The MDM bases the estimation on its information about the current capacity of the vTree. If there is insufficient capacity at the destination based on that estimate, migration does not start. (An advanced option allows you to force the migration even if there is insufficient capacity at the destination, with the intention to increase the capacity as required during the migration.) The MDM does not reserve the estimated capacity at the destination (since the capacity of the source volume can grow during migration and the reserved capacity does not guarantee success). The MDM does not retain source capacity once it has been migrated, but releases it immediately.

Use the following table to understand which vTree migrations are possible, and under what specific conditions:

When migrating from a medium granularity storage pool to a fine granularity storage pool, volumes must be zero padded.

You can pause a vTree migration at any time, in the following ways:

• Gracefully—to allow all data blocks currently being migrated to finish before pausing.
• Forcefully—to stop the migration of all blocks currently in progress.

Once paused, you can choose to either resume the vTree migration, or to roll back the migration and have all volume blocks returned to the original storage pool.

vTree migration might also be paused internally by the system. System pauses happen when a rebuild operation begins at either the source or destination storage pool. If the migration is paused due to a rebuild operation, it remains paused until the rebuild ends. If the system encounters a communication error that prevents the migration from proceeding, it pauses the migration and periodically tries to resume it. After a configurable number of attempts to resume the migration, the migration remains paused, and no additional retries will be attempted. You can manually resume migrations that were internally paused by the system.

Concurrent vTree migrations are allowed in the system. These migrations are prioritized by the order in which they were invoked, or by manually assigning the migration to the head or the tail of the migration queue. You can update the priority of a migration while it is being run. The system strives to adhere to the priority set by the user, but it is possible that volume blocks belonging to migrations lower in priority are run before ones that are higher in priority. This can happen when a storage pool that is involved in migrating a higher priority block is busy with other incoming migrations, and the storage pools involved in lower priority migrations are available to run the migration.