NDMP access to object replication target file systems

Replication and Disaster Recovery Administration Guide for Hitachi NAS Platform

Part Number

Read only access to a target file system is possible using the NDMP protocol, but note that some restrictions apply:

  • Only backups are allowed.
  • Snapshot Rules must be used.
  • Object replications must use snapshot rules at the target.
  • Setting the snapshot rule queue depth is critical.
  • Snapshot rules for backups use Hitachi Vantara-specific settings.

The following areas of functionality are affected by the file system being an object replication target:

  • Incremental backup mechanisms.
  • Direct access recovery.
  • Rollback of the object replication target.

The restrictions and considerations are explained in detail here:

Only backups are allowed

When using NDMP, only backups work on a replication target. Besides backups, other NDMP operations need to be able to write to the file system, for example file replications need to be able to create snapshots. All other NDMP operations besides backups will fail on an object replication target. Recovery from backed up data would be made to a different file system, or the object replication target would need to be promoted. The backup need not be of the whole file system, and sub-directories or virtual volumes can be backed up normally.

Snapshot rules must be used

The file system state of a replication target is created only by the object replication, and nothing else. NDMP backups must reference a past state of the file system in order to perform incremental backups. A snapshot is created by the snapshot rule defined for the replication policy. There is an existing mechanism by which NDMP operations can already specify a snapshot rule. The latest snapshot in the specified rule becomes the dataset that is backed up. Only backups that specify a snapshot rule are allowed. All others will fail.

Setting the snapshot rule queue depth is critical

If more object replications are run using the snapshot rule than the queue depth, the oldest snapshot in the queue is destroyed. If the backups are run less frequently than the replications, it is possible that the destroyed snapshot was one that the incremental backups were using as a base. The frequency and timing of the object replications and backups must be very carefully planned so the base snapshot is not destroyed.

Incremental backup mechanisms

There are two ways that NDMP backups can specify the base from which an incremental backup is made:

  1. The standard practice of using backup levels.
  2. Use token-based backups.

Both work correctly as long as a snapshot rule is specified. The level or token defines the base file system state that the incremental backup refers back to and the snapshot rule defines the state of the file system at the point of backup.

Direct Access Recovery

Although it is not possible to recover to an object replication target, you can use backups made of the target to recover to a different file system, via Direct Access Recovery (DAR). Per typical practices, the NDMP environment variables must be set to support such a recovery.

Rollback of the object replication target

Rollback of the object replication target affects incremental backups. No data is lost, but backups that immediately follow the rollback must re-establish themselves to lower levels, if they depended on a snapshot that is lost in the rollback. This could cause the backup to take longer than might normally be expected.

Note: Specifying snapshot rules for backups is a Hitachi Vantara proprietary setting. The existing mechanism for specifying a snapshot rule is to set the NDMP environment variable NDMP_BLUEARC_USE_SNAPSHOT_RULE to the name of the rule that you wish to use. This is not possible on all backup applications. To permit the use of backup applications where it isn't possible, use ndmp-backup-snapshot-rules. This command allows the per-file system mapping of a snapshot rule to a particular level of backup. Allowing the use of different snapshot rules for different levels allows more flexibility in setting the queue size of the rules e.g. low numbered backup levels happen rarely and require their snapshots to be kept for a long time

Please see the NAS man pages for more detailed information on using commands for backups and snapshots.