Hi,
So, I must be misunderstanding something about how this works.
At the moment my knowledge is the following:
* If everything is fine, each hyper-v host in the cluster connect DIRECTLY to the storage - traffic is: Host <-> Storage
* If a host loses its connectivity to the storage, then it will redirect the traffic to another host that still has access to the storage. Some some traffic is Host <-> Storage, and some traffic is Fail Host <-> Host <-> Storage
----
Now, at this specific moment, I have a 2 node Hyper-v 2012 R2 cluster connected to an iSCSI. Hosts are using MPIO and I am somewhat multiply connected to the SAN
The FCM reports that the CSV is not in redirected access. However, only one hosts shows the CSV disk in Nodes -> Disks.
----
Now, on Node A, I am running a VM on which I am running SQLIO in preparation for SQL Server install. Node A has 2 Ethernet cards 10 GB on the iSCSI network. Node B has the same. Node A is the one not showing the CSV disk in Node A -> Disks. This is what
I see: (reading test, not yet up to the writing ones)
Node A - Eth1 - S:5 Mbps, R:900 Mbps
Node A - Eth2 - S: 5Mbps, R:900 Mbps
Node B - Eth1 - S: 10Mbps, R:900 Mbps
Node B - Eth2 - S: 1.8 GIGA bps, R:900 Mbps
----
Writing test:
Node A - Eth1 - S: 500 Mbps, R:2 Mbps
Node A - Eth2 - S: 500 Mbps, R:2 Mbps
Node B - Eth1 - S: 1.4 GIGA bps, R:9 Mbps
Node B - Eth2 - S: 800 Mbps, R:1.1 GIGA bps
----
So to me this means:
* the traffic from the VM is actually routed through the other node.
* Node B is connected to both storage controller but only 1 port on each. (not the issue atm, but eventually I'd like to see all 4 ports)
* Node A is connected to Node B
* Node A and Node B MPIO works very well.
----
When it is the writing test:
* Node B is sending twice the data because of the 2 way mirroring (I am using tiered storage and it is configured on the hyper-v)
----
As for the storage, it sticks at 220 MB/sec for reads or writes at the moment, which is normal considering that only 2 out of the 8 Ethernet ports are used (1Gbps)
Edit: It is actually starting to use more than 2 out of 8. Write went up to 360 Mbps
----
Question: Am I misunderstanding something when it comes to hyper-v cluster and redirected access ? If yes, what is it and how do I fix it? (Or where is the documentation on how to fix it)
Thanks
Edit: Seems it is linked to the fact that by default when you use tiering it enables heatmap on the volume which causes the redirection