Alletra MP with iSCSI hosts

[CadenLange]

Hi all

In an environment whereby each Alletra MP node has two configured iSCSI ports on each node (0:4:1; 0:4:2; 1:4;1, 1:4:2) configured with two iSCSI vlans - ports 0:4:1 and 1:4:1 are on iSCSI VLAN 1 and ports 0:4:2 and 1:4:2 are on iSCSI VLAN 2.

How many 'Active (I/O)' paths should an ESXi host see? 

As the Alletra MP sees 4 paths to the ESXi host with RR multipathing and with the array being active\active should it see active IO on all 4 paths or not?

many thanks

[support_s]

System recommended content:

1. HPE GreenLake for Block Storage: VMware ESXi Implementation Guide | Configuring a jumbo frame on iSCSI ports on the storage system

2. HPE GreenLake for Block Storage: VMware ESXi Implementation Guide | Configuring the host for a Software iSCSI connection

Please click on "Thumbs Up/Kudo" icon to give a "Kudo".

Thank you for being a HPE valuable community member.

[Satish04]

Hi CadenLange,

When using HPE Alletra MP arrays and Round Robin (RR) multipathing, an ESXi host should see four active I/O paths under normal circumstances.
Using VMware's Round Robin policy, I/O requests are distributed across all active paths, alternating between them.
When using an active/active configured HPE Alletra MP array, all four iSCSI paths are available to each ESXi host across both VLANs.
The Round Robin multipathing mechanism ensures that I/O is balanced across the 4 paths if there are no network or configuration issues.

Hope this helps.!

Regards,
Satish

[Cali · Accepted solution]

@Satish04 

Are you sure?
This was the case with 3PAR/Primera, but with Alletra Block there is a new Path Handling:

A host’s IO paths to a VV will be ALUA active “optimized” to the node that owns the VV and will be ALUA active “non-optimized” to all other nodes in the array.
This is consistent with my experience that there are 4 (8 with mirroring) active paths, but only two are active (I/O) if there are two host connections per node.

In this Picture on Array xx57 Node 0 with Port 0 & 1 and not on Node1.
This reduces Proxy Reads between the Nodes.

Cali

[support_s]

Hello,

Let us know if you were able to resolve the issue.

If you have no further query, and you are satisfied with the answer then kindly mark the topic as Solved so that it is helpful for all community members.

Please click on "Thumbs Up/Kudo" icon to give a "Kudo".

Thank you for being a HPE valuable community member.