What are the best practices for designing SnapMirror to minimize user impact when using dedicated ports e0d/e0f ?
Applies to
ONTAP 9
Answer
Best practices and design recommendations:
- Separate SnapMirror and user traffic
- Use dedicated physical ports (e0d/e0f) for SnapMirror intercluster traffic.
- Prevents replication traffic from affecting user data access.
- If possible, use separate VLANs or network segments for further isolation.
- Create dedicated intercluster LIFs for SnapMirror
- Create at least one intercluster LIF per node; two per node (e0d, e0f) recommended for redundancy.
- Assign intercluster LIFs only to e0d/e0f, not to ports used by user traffic.
- Do not include intercluster LIFs in the failover group for data (e0c/e0e).
- Intercluster LIF failover is node-local; enable auto-revert so LIFs return to their home port after recovery.
- SnapMirror policy and schedule
- Asynchronous SnapMirror is suitable for disaster recovery if recovery point objective (RPO) is acceptable.
- Schedule SnapMirror jobs outside business peak hours (e.g., early morning, evening).
- Network bandwidth control (throttling)
- To prevent SnapMirror from saturating the network, configure bandwidth limits:
- Global throttling: system-wide limit for all SnapMirror transfers.
- Relationship-level throttling: set per SnapMirror relationship using
-throttle <KB/s>on snapmirror create or snapmirror modify.
- Reference: How to set up SnapMirror replication throttling for system and individual transfers
- To prevent SnapMirror from saturating the network, configure bandwidth limits:
- Additional notes
- Management ports are not supported for SnapMirror intercluster LIFs; use data ports (including e0d/e0f).
- Intercluster LIFs can share physical ports with data LIFs, but dedicated ports are preferred for isolation.
- For further details on networking requirements, see SnapMirror configuration and best practices guide for ONTAP 9 (see “Intercluster networking requirement”).
- Cautions
- Intercluster LIFs only fail over within the same node; plan redundancy accordingly.
- Do not mix intercluster and data LIFs in the same failover group.
- Confirm switch port availability and connectivity for e0d/e0f before deployment.
