[ONTAP] Convert to SVM-DR Research

A page for research/notes/lessons-learned on converting a primary/DR SVM pair from volume snapmirror relationships to an SVM-DR (or SnapMirror for SVM) relationship.

Notes from the above:
  • Each volume (except SVM root) is being replicated
  • Each volume on the source (including SVM root) has same name as on destination
    • use volume rename when SnapMirror is idle
    • use snapmirror resync after rename
  • Create SVM replication as per: Replicate an entire SVM configuration
    • snapmirror create ... -identity-preserve true
  • Stop destination SVM: vserver stop -vserver SVMB
  • snapmirror resync -source-path SVMA: -destination-path SVMB: -type DP|XDP -schedule SCHEDULE -policy POLICY

Q: What does identity-preserve preserve?
A: See: About SnapMirror SVM replication #configurations-replicated-in-svm-dr-relationships

Q: DP or XDP?
A: XDP (it is the default from ONTAP 9.4)

Q: What policy options?
A: async-mirror (SnapMirror DR) or mirror-vault (Unified Replication)

Q: Can you SVM-DR Flexgroups?

Q: What about SnapMirror Business Continuity?
A: NetApp SnapMirror Business Continuity (SM-BC) is a continuously available storage. solution with application-level granularity  Note: SnapMirror Business Continuity offers application-level granularity for SAN workloads only.

Notes from #configurations-replicated-in-svm-dr-relationships:

The key thing is that SVM-DR isn't super useful for SAN (still useful for replicating an entire SVM for DR purpose) as we see if we look at the things that are not replicated with -identity-preserve true (updated 11/11/2022):
  • Not replicated with -identity-preserve true:
    • Network: SAN LIFs, Broadcast domain, Subnet, IPspace
    • Root volume: User data, Qtrees, Quotas, File-level QoS, Attributes
    • Fibre Channel (FC): No
    • iSCSI: No
    • LUNs: igroups, portsets, Serial numbers
Other reading

Older Stuff On Cosonok's IT Blog
  • Prerequisites for cluster peering
    • Key thing is that each node in the cluster (source and destination clusters) must have an intercluster IP address. SnapMirror requires a full mesh to work (so if you have 2 nodes at the source, 2 nodes at the destination, you need minimum 4 IPs and routing from source to destination cluster and vice versa.)
  • A bit on configuring SVM DR using the GUI is covered here: