Enabling iSCSI on azure local

Question

Enabling iSCSI on azure local

Daniele Francioni 21

Hello

We have a 6 node azure local installation configured in hyperconvergence mode (the default installation). We would like to enable iSCSI as per (https://learn.microsoft.com/en-us/azure/azure-local/deploy/enable-external-storage?view=azloc-2606#prerequisites). Our nodes have 2 mellanox 25 gbps fiber nic that are used for all intents [network, compute, storage]. The instructions states to move the storage intent to a new set of nics (vNics are not supported) that should be reserved for iSCSI communication. Some questions:

does all storage traffic (hyperconvergence replicas) be directed to the new nics?
is there a way to reserve the new nics only for iSCSI traffic, avoiding to change the network intent or to redirect the storage traffic?
since the new nics are 1 gbps, should I prefer to mantain storage intent on the mellanox (25 gbps) nics and move all other traffics on the new nics? However I prefer to not change the current intent configuration, if possibile.

0 comments

Answer accepted by question author

Anonymous

Daniele Francioni

iSCSI does not consume or relocate your S2D storage intent. The requirement is to add separate, dedicated physical ports for iSCSI & not to move your existing S2D (hyperconverged replica) traffic. - https://learn.microsoft.com/en-us/azure/azure-local/deploy/enable-external-storage?view=azloc-2606

Does all storage traffic (S2D replicas) move to the new NICs?

No. Your S2D/SMB replica traffic should stay on the 25 GbE Mellanox storage intent. iSCSI is a separate block-storage path to the external SAN and requires its own dedicated physical ports and it does not redirect or absorb your S2D east-west replica traffic - https://learn.microsoft.com/en-us/azure/azure-local/deploy/enable-external-storage?view=azloc-2606

Can the new NICs be reserved only for iSCSI without changing the network intent?

Yes that is exactly the supported design. The dedicated iSCSI ports are not added to any Network ATC intent. You leave your existing converged (mgmt/compute/storage) intent untouched, and configure the new physical NICs outside Network ATC purely for iSCSI + MPIO connectivity to the SAN. Requirement: dedicated physical ports (vNICs are not supported), identical NIC config across all 6 nodes, and cluster on version 2604 or later.

Should you keep S2D on the 25 GbE Mellanox and move other traffic to the 1 GbE NICs?

Keep S2D on the 25 GbE Mellanox NICs — do not move the storage intent. However, be cautious: using 1 GbE ports for iSCSI is a likely bottleneck for SAN block I/O and is generally not advisable for production data workloads. The recommended approach is:

Mellanox 25 GbE and keep current converged intent (incl. S2D storage). - https://learn.microsoft.com/en-us/azure/azure-local/deploy/enable-external-storage?view=azloc-2606

Dedicated NICs (ideally ≥10/25 GbE) and reserved exclusively for iSCSI/MPIO to the SAN.

If 1 GbE is your only option for the extra ports, it will work functionally for the preview, but plan for higher-speed dedicated iSCSI NICs to avoid performance limits.

0 comments

1 additional answer

Your answer

Answer 1

Thank you for your answer. my doubts were related to this point of the original document

https://learn.microsoft.com/en-us/azure/azure-local/deploy/enable-external-storage?view=azloc-2606#31-exclude-iscsi-nics-from-network-atc

Add-NetIntent -Name "Mgmt-Compute" -Management -Compute -AdapterName "NIC1","NIC2"

Add-NetIntent -Name "Storage" -Storage -AdapterName "NIC3","NIC4"

These commands seems to change the ATC so that the Mgmt-Compute will use only the NIC1 and 2 (mellanox, in my case), while the storage is used only by NIC3 and 4 (gbps). In my scenario, since the two NICs are of different types, i should not even update the storage network intent to include also these two, right?

I understand that 1 gbps could be a bottleneck, but in our scenario is acceptable because it is used as archive storage by our workloads rather than hot storage.