Views:

Summary



The NetApp FlexClone license is a feature that DPX integrates with to simplify recovery and maintenance operations. Catalogic Software recommends that DPX customers license their NetApp devices with FlexClone to enjoy additional conveniences with managing the data protection solution.

Resolution



Introduction

NetApp FlexClone is a feature that enables ONTAP to create volume copies with little or no additional disk or CPU overhead on the system. You can read and modify these copies without affecting the original source volume in any way. The source volume can be read/write or read-only, and the lightweight FlexClone copy is read/write. It is helpful to use these lightweight copies of volume data for testing, development, or data recovery purposes, especially when modifying the original source volume is not allowed or impossible.

For additional details about NetApp's FlexClone license, features, and use, please see the "Virtual Data Replication with FlexClone Technology" chapter in NetApp's Data ONTAP 8.0 Storage Efficiency Management Guide.

General DPX Interaction Summary

While FlexClone licensing does not have any effect on DPX backups, DPX does interface with FlexClone for restore and space recovery operations, which is an especially important function for environments that replicate data using SnapMirror technology.

DPX supports a number of advanced features that utilize NetApp's iSCSI LUN facilities:

  • Instant Availability (IA)
  • Instant Virtualization (IV)
  • Rapid Return to Production (RRP)
  • Bare Metal Recovery (BMR)
  • On-Demand Verification
  • Application Verification

You create iSCSI LUNs by accessing data from a point-in-time snapshot and writing changes to a LUN file that resides at the root of the volume for which the snapshot exists. Therefore, the requirement for using iSCSI with DPX depends on a snapshot and a read/write space within that snapshot's volume.

Space Management

With FlexClone license, any operation that needs to create an iSCSI connection will first rapidly create a lightweight volume clone of the data, and then proceed with all read/write operations referencing that cloned data. From this respect, there is no access made to the original data volume; all data access is isolated to the cloned volume.

Without the FlexClone license, DPX must use the original data volume to create iSCSI LUN files. If you use the LUN extensively, access to the LUN might overlap backup processes scheduled for that volume. This includes scheduled DPX backups to that volume, and volume snapshots scheduled directly through ONTAP.

LUN data that is active at the time of any snapshot (Including DPX backups) will retain that data within the snapshot until you delete the snapshot. This may not be desirable, and may consume unnecessary space until you delete all snapshots with that LUN.

A snapshot created on a volume with an active LUN, the snapshot from which you created the LUN will remain in a "LUN-busy" state. Each successive snapshot will build dependencies on previous snapshots, creating a chain of LUN-busy snapshots that you cannot remove easily.

When snapshots are stuck in "LUN-busy," the DPX condense process will not be able to remove these snapshots automatically. The condense process will fail for that volume, and the data (which should be expired) will not be removed. Thus, the unique data that snapshot contains will not be freed from the volume.

You can mitigate this condition with a per-volume option setting that will relax the LUN dependencies, and allow the deletion of these snapshots. This option must be set manually on each volume at the time of creation for the volume. If the option is not set ahead of time, LUN-busy snapshots can accumulate, and the administrator will need to enable the option, and then re-run the condense process to clean up the expired data.

One other consideration is deduplication use within the original volume. If the original volume is used for iSCSI LUN creation and backup operations occur and/or scheduled deduplication (ASIS) process kick off, the extra LUN data will consume additional CPU time and space to complete the deduplication process. The LUN may not have anything that deduplication can save, and the resulting point in time snapshot will needlessly save LUN data (consuming space) that isn't necessary.

Please also note that the more servers you have in a job going into a volume, the greater the likelihood that these conditions will occur. IA is a frequently used and convenient feature for administrators to use when recovering data. Other features like IV, FV, RRP, and BMR will have similar affects. As the number of servers in a volume increases, the importance of that job running also increases, as does the likelihood that one or more restore operations will be in progress at the time the backup job runs again. The more servers that are backed up to a single volume, the greater the importance a FlexClone license plays. Moving temporary data and I/O to a lightweight volume clone saves space by preventing the trapping of this data in the snapshots of the original volume.

Advanced Recovery Operations with SnapMirror Replication

SnapMirror is a popular feature used by DPX customers to replicate data off-site. The source volume remains in a read/write state, and the destination volume remains in a read-only state. The SnapMirror destination volume is always in a read-only state to permit read access to the data, but to allow the source to update it as needed.

DPX allows for the data recovery from an 'alternate secondary' location. The alternate secondary is typically a SnapMirror destination volume or the restore of a SMTape data set (although, it is possible to copy/restore data in other ways for alternate secondary use).

When the volume is marked read-only, you cannot use that volume to create iSCSI LUNs. You must create a LUN on a volume that is in a read/write state.

FlexClone rapidly creates a volume copy in a read/write state that it uses for iSCSI LUN operations. Again, the LUN data is completely contained within this clone separate from the original volume, and with no additional user intervention, DPX read, write, or destroyed it at any time.

When FlexClone is not licensed, DPX must use the original volume to create the iSCSI LUN. If the volume is in a read-only state, the data must be available in a read/write state for the LUN operations to work. The NetApp administrator has one of two options in this case:

1) Manually break off the SnapMirror relationship. The volume will then be in a read/write state for LUN usage. When you no longer need the LUN, resync SnapMirror to re-establish the SnapMirror relationship. If the volume has been broken-off for an extended period, the source volume can accumulate changes that would make re-synchronization difficult (bandwidth, time, etc.). LUNs containing very large volumes for BMR, RRP, or are used for IV's that need to be active, may have to leave the volume broken-off for an extended period. If you improperly delete volume snapshots from the copy, this might require a full SnapMirror transfer or initialization to correct.

2) Copy the data to another read/write volume. This can work in theory, but is usually not a viable solution. It can take a very long time to copy the data, and this will affect how quickly you can access the restore sets you need. A local NetApp system may not have sufficient space to copy the data, and transfer of such across the network to another NetApp device may not be workable.

FlexClone also helps to prevent unnecessary replication of LUN data to your off-site NetApp device. With FlexClone enabled, a clone of the source volume can create iSCSI LUNs for DPX features. LUNI/O activity is isolated to the cloned volume. Without FlexClone, create the iSCSI LUN in the original volume. SnapMirror will then replicate all of that LUN I/O to the remote NetApp system. LUNs created for IA, IV, and RRP that need to remain active for long periods of time, will see all of the LUN changes replicated off-site, and this may affect bandwidth use between sites, remote storage capacity, etc.

As indicated in the previous section, multiple servers backing up to a single volume increase the potential for challenges to occur. The SnapVault secondary NetApp device will have more frequent restore requests and higher likelihood for several concurrent restore requests with volumes holding multiple server backups. This will not only trap unnecessary data in new backup snapshots, but SnapMirror will also replicate that volume data off-site. Using "alternate secondary" features and BMR from the SnapMirror destination volume will also introduce challenges. The greater the number of outstanding concurrent restore requests, the longer the SnapMirror destination volume will need to remain broken-off, and this will increase the backlog of SnapMirror data that needs to be transmitted.

FlexClone license resolves both of those challenges. On the SnapVault secondary, FlexClone will keep all operations to a local lightweight volume copy that it does not automatically replicate off site. On the SnapMirror destination device, the volume never needs to be broken off, leaving the replication schedule undisturbed and eliminating manual intervention.