NVMe Namespace Does Not Support Selective LUN Mapping in ONTAP
Applies to
- NetApp ONTAP 9.5 and later
- All Flash FAS (AFF) systems
- NVMe-oF (NVMe/FC, NVMe/TCP) protocols
- VMware ESXi hosts
Answer
- This is expected behavior. ONTAP does not support Selective LUN Mapping (SLM) for NVMe namespaces. All NVMe data LIFs zoned to a host are presented as available paths for each NVMe namespace.
- Unlike SCSI/FC protocols, SLM is not available or required for NVMe, as NVMe pathing and performance rely on queue-pair parallelism rather than path count.
- Explanation
- ONTAP intentionally limits NVMe data LIFs to 2 per SVM per node to control overall path scale.
- All NVMe LIFs zoned to the host are visible as active paths for the namespace.
- NVMe performance is not impacted by fewer paths; high IOPS are achieved through NVMe queue parallelism.
- This is not a bug or misconfiguration. This behavior is by design and not a software issue.
- Best Practices to Manage Path Count
- Zone only the minimum required NVMe LIFs to each ESXi host (ideally 2 per SVM per node).
- Avoid zoning all NVMe LIFs from all nodes unless required for redundancy.
- If additional LIFs are required for load balancing, create additional SVMs (each SVM supports 2 NVMe LIFs per node).
- Note:NVMe performance is not dependent on path count. It uses a queue‑pair model for parallelism, allowing high IOPS with minimal paths, unlike SCSI/FC which may require multiple paths for performance.
Additional Information
