Redirecting Inter-Controller Communication on SV3200 4-port 1GbE iSCSI Storage systems

[Fred Blum]

Our SV3200 has the traffic type for Storevirtual OS traffic set to Bond0. This was the default during setup. 

Is there a way for isolating the Storevirtual OS traffic? Seperating it from the iscsci data channels?

We are using StoreVirtual 3200 4-port 1GbE iSCSI Storage systems and I need it to further pinpoint the package drops issue.

Thanks in advance.

Fred

[Fiberfish]

Hi,

In SV3200 , Bond0 is formed with Eth3 and Eth4 ports bonded together in four port ICSI setup.

and Eth0 and Eth1 are disabled by default and advised to remain disabled , When Eth3 and Eth4 are present.

Eth2 is used for Management traffic. We can put Eth2 in a different Subnet than that of Bond0 ( Eth3 and Eth4) and if you are using FOM you are use Router for communicating between two different Subnet.

Regards

I am an HPE Employee

[Fred Blum]

The SV3200 we have are 2 node storage enclusures with 2 storage controllers and 4 iScsi Port.

That is 2 per Storage controller. Default those are bound together, so Eth1 and Eth2 are Bond) on SC1 and Eth3 and Eth4 are Bond0 on SC2. They still use 4 IP adresses for ALB so in an optimal situation as per HPE installation manual you are requested to create 4 volumes or even 8 as the volumes are assigned to the respective IP adresses with ALB. 

They only way possible is to break the bond an dedicate a NIC to iscsi and one to SV OS traffic per SC sacrificing ALB.

[Fiberfish]

Hi,

Thus, you have a single storage SV3200 array and thus with two controllers in single enclosure.

This have defualt " Eth0 " , " Eth1" , two ISCSI data port per controller. In both the controller say SC1 and SC2, they are called Eth0 and Eth1 only. There will no Eth3 and Eth4 for you, that is"  only available if you have four ISCSI data port per controller.

Thus, if you download SV3200 support bundle and see "ifconfig.log " file there you will see Eth0 and Eth1 which are bonded together to form Bond0 and by internal design of the array, Bond0 is meant for carrying data traffic only .

For Management traffic ( Management of the array ) you need to use Eth2 on the array. 

We use two IP address per controller , thus four IP address as we have two controller. This is required for load balancing .

Also, we have two transport_server per controller , thus four transport_server in two controller, that is the reason for optimal use of bandwidth available in controllers it is suggested that we distribute load on four or multiple of four number of volumes.

and it is advised we follow the default setup, Removing Bond0 is not a suggested action in SV3200 as this is the way array architecture works.

Regards

I am HPE Employee