MicrosoftCosmos_Darwin I am thankful that you explictly state the one room / multi room deployments in this exciting announcement.
In the past asked whether they may deploy an Azure Stack HCI across different rooms, in a special scenario, without the given complexity of a Streched Cluster.
I have just downloaded the recent documentation for Azure Stack HCI (en-us). There are only 3 times you can find the word "room" or "same rack"
Citing Azure Stack HCI on learn.microsoft.com:
Streched Cluster requirements:
"The sites can be in different countries/regions, different cities, different floors, or different rooms."
Networking requirements:
"We highly recommend that all cluster nodes in a site are physically located in the
same rack and connected to the same top-of-rack (ToR) switches."
I would like to know your opinion on whether it is helpful adding details in the documentation about a certain frequently asked scenario:
An Azure Stack HCI cluster consisting of two nodes of version 22H2 or 23H2. using Network ATC.
The storage network is directly connected to each of the nodes via fiber (L1), using RoCe v2 (RDMA)
The compute / management network is switched and does not use RDMA. They do not use ToR switches because they do not need them, but they have two backup core layer switches in each room.
Each node! of the two nodes is located in a different room to protect them from fire or other disasters.
The documentation states it's "highly recommended" not putting them apart.
sBut at the same time doesn't seem to consider an unswitched direct connection.
So the question of customers is centric about the downsides and risks, and whether this scenario is supported, given the citations.
Thank you very much on for your help on this matter!
In addition, are there plans that the deployment of streched clusters will be automagically done as we see it for two nodes scenarios starting with this release?
This would be pretty awesome.
