3 Node HA Network configuration

[Slingshotz]

I'm having problems connecting the iSCSI targets on a new v8 installation and wondering if what I have designed should work. In v6, there were 3 sync channels on the configuration manual diagrams but in v8 there seems to be only 2 sync channels. We have three servers, each with two 10Gig ports that I've connected directly to each other to bypass the need for a 10gig switch. The 192.168.7.x network is only for heartbeat. I've created the CSVs fine with replication and the iSCSI discovery is working but when I am trying to connect the target in MPIO I am unable to choose the target portal IP that is connected to the initiator IP. From following the beta manual, on server 1 I should be able to choose 10.1.2.2 as a target portal IP when I choose 10.1.2.1 as the initiator IP but I do not get that option.

Design.JPG (24.26 KiB) Viewed 10271 times

iSCSI_Discovery.JPG (28.45 KiB) Viewed 10270 times

iSCSI_TargetMPIO.JPG (44.19 KiB) Viewed 10270 times

[anton (staff)]

What hypervisor are you using: Hyper-V or vSphere? From what I understand you don't run hyper converged, rather you use "old school" dedicated setup. Is this correct?

[Slingshotz]

We are running Hyper-V but I do not understand what you mean by converged or dedicated setup.

[anton (staff)]

Converged. StarWind Virtual SAN and Hyper-V run on the same hardware. iSCSI is used to establish an inside-node connection, then traffic turns to go over our loopback accelerator driver (fast path). Microsoft does similar things when they go from TCP -> RDMA with SMB 3.0 protocol. Connect starts to live as TCP and then it's RDMA. V8 + Windows Server 2012 R2 setup is different here compared to V6 Native SAN product (no loopback acceleration there, iSCSI was used).

Dedicated. StarWind Virtual SAN and Hyper-V run on the different hardware (two separate layers, one compute layer and one storage layer). SMB 3.0 is used to talk from Hyper-V to storage layer (Scale-Out File Servers). StarWind Virtual SAN supports shared block back end for SoFS config. This setup was experimental with V6 but with V8 it's a primary one.

We are running Hyper-V but I do not understand what you mean by converged or dedicated setup.

[Slingshotz]

Ahhh, yes we are running converged where they both are on the same piece of hardware. Now that I look at the network design, we don't really have any MPIO as there is technically only one way that Server 1 can connect to Server 3. Is that still a viable configuration or would I need to configure some kind of network routing to allow a connection from Server 1 to Server 3 through Server 2?

[Anatoly (staff)]

Hi!

Just a quick question - I see that you have valid ASM. Do you mind taking this conversation to the email/phone conversation and posting the results (if you want) here after?