VPLEX: Virtual Microsoft Failover Cluster storage validation report fails for Distributed Devices
Sammanfattning: If a Windows Server Failover Cluster is hosted on a ESXi virtualization host and uses VPLEX Metro Distributed Devices as physical RDMs to the cluster VMs, the cluster storage validation test in the VM may fail if the NMP selected paths point to different VPLEX clusters. This is expected behavior. Cluster disk functionality is not affected. ...
Den här artikeln gäller för
Den här artikeln gäller inte för
Den här artikeln är inte kopplad till någon specifik produkt.
Alla produktversioner identifieras inte i den här artikeln.
Symptom
Given the following environment:
- A Windows Server Failover Cluster,
- formed by virtual machines on a vSphere ESXi hypervisor,
- using VPLEX Metro Distributed Devices,
- provisioned as physical RDMs to the cluster virtual machines,
- using NMP VMW_SATP_INV with VMW_PSP_FIXED.
A notable difference lies in the VPD serial number section.
When the storage tests succeed, they show identical serial numbers.
When they fail, the serial number is different.
Device failover functionality is not affected, but the validation results vary. This is turn may raise supportability concerns.
Orsak
In the above virtual cluster scenario, the actively selected path of the hypervisor-level multipath layer (NMP) may at times point to different VPLEX clusters.
If this happens, the SCSI inquiry sent to the corresponding RDM disk object in the virtual machine may return a different serial number for the same Distributed Device presentation as on another cluster node.
Upplösning
The virtualized cluster behaves like a physical one.
In the above virtual cluster scenario, the actively selected path of the hypervisor-level multipath layer (NMP) may at times point to different VPLEX clusters.
If this happens, the SCSI inquiry sent to the corresponding RDM disk object in the virtual machine may correctly return a different serial number for the same Distributed Device presentation compared to a peer cluster node.
This is expected behavior.
Cluster disk functionality is not affected. This is a validation report symptom only.
The behavior is documented in the Dell Host Connectivity Guide For Windows in the Windows Failover Clustering with VPLEX section.
The guide can be accessed on the Dell Support page at https://elabnavigator.dell.com/eln/hostConnectivity.
The behavior is also further explained in detail on Microsoft web page https://learn.microsoft.com/en-us/troubleshoot/windows-server/high-availability/failover-cluster-storage-validation-tests-dont-discover-all-shared-luns.
In the above virtual cluster scenario, the actively selected path of the hypervisor-level multipath layer (NMP) may at times point to different VPLEX clusters.
If this happens, the SCSI inquiry sent to the corresponding RDM disk object in the virtual machine may correctly return a different serial number for the same Distributed Device presentation compared to a peer cluster node.
This is expected behavior.
Cluster disk functionality is not affected. This is a validation report symptom only.
The behavior is documented in the Dell Host Connectivity Guide For Windows in the Windows Failover Clustering with VPLEX section.
The guide can be accessed on the Dell Support page at https://elabnavigator.dell.com/eln/hostConnectivity.
The behavior is also further explained in detail on Microsoft web page https://learn.microsoft.com/en-us/troubleshoot/windows-server/high-availability/failover-cluster-storage-validation-tests-dont-discover-all-shared-luns.
Ytterligare information
Remarks
- Older versions of the Host Connectivity Guide For Windows refer this behavior to "Windows Server 2012" and "Windows Server 2012 R2". This behavior applies to all Windows Server releases, as is also indicated in the Microsoft solution link (below). The Guide has been modified.
- Older versions of the Host Connectivity Guide For Windows may mention the now deprecated Microsoft KB 2914974. Microsoft has republished the content at https://learn.microsoft.com/en-us/troubleshoot/windows-server/high-availability/failover-cluster-storage-validation-tests-dont-discover-all-shared-luns
- The behavior is independent from the vSphere ESXi NMP path selection policy. It will also happen with round-robin.
- The behavior is rooted in a principle behavior of VPLEX-Distributed Devices. It is therefore independent of the hypervisor used. Similar results are expected with Microsoft Hyper-V, and others.
Berörda produkter
VPLEXArtikelegenskaper
Artikelnummer: 000221583
Artikeltyp: Solution
Senast ändrad: 05 apr. 2024
Version: 5
Få svar på dina frågor från andra Dell-användare
Supporttjänster
Kontrollera om din enhet omfattas av supporttjänster.