After fielding a number of questions around this topic recently, its come to my attention that there is a misunderstanding on how to implement the vSphere Storage Appliance (VSA) for certain ROBO solutions. The requirement around the tie-breaker code (VSA Cluster Service) is either misunderstood or misinterpreted when it comes to 2 node VSA deployments at remote office/branch office (ROBO) sites.
Let’s be clear about it now – the VSA needs 3 votes in order to function correctly as a cluster. In a 3 node VSA deployment, this is not an issue since each node gets a vote. In a 2 node VSA deployment, the third vote is handled by the VSA Cluster Service aka the tie-breaker code.
Now, in standard (non ROBO) 2 node VSA deployments where the vCenter is located in the same datacenter as the VSA, the VSA Cluster Service can exist on the vCenter server. During VSA installation, you simply provide the IP address of the vCenter server for the VSA Cluster Service.
However, in ROBO deployments, things are a little different. In these deployments, the vCenter server is typically located at a central office, and the 2 node VSA is deployed at a remote office. The third vote (tie-breaker code) must also be located at the remote office/branch office. To cater for this, we allow the VSA Cluster Service to be decoupled from vCenter and installed in a VM. The VM can be running on a non-VSA ESXi host or Workstation at the branch office. We cannot have it run on the ESXi hosts involved in the VSA cluster because if we lost the host on which the tie-breaker VM was running, we lose 2 votes, and thus the cluster quorum. So it must exist outside of the VSA.
VMware provides VSA Cluster Service installation documentation for the different platforms that we support. Currently there is Standalone VSA Cluster Service for Linux and a Standalone VSA Cluster Service for Windows. The VSA Cluster Service platform must be on the same subnet as the VSA cluster members.During the install, then installer will ask for details of which host has the tie-breaker code, so the VSA Cluster Service must be installed first. The VSA Installation wizard will then audit this, making sure that it can reach and communicate to the VSA Cluster Service code.
Please make sure you take into account the requirement for the VSA Cluster Service at the remote office/branch office site when designing or achitecting 2 node VSA ROBO solutions.
Get notification of these blogs postings and more VMware Storage information by following me on Twitter: @VMwareStorage
Thanks Cormac for the info. Tricky situation indeed when the 2 boxes are completely isolated and the entire office only has these 2 boxes. Looks like it has to be a “server” (low end tower server machine). In future, can it be a network switch or a cloud service (assuming the machine can connect to internet)? Thanks from Singapore.
Hi Iwan, we are looking at some other options for the quorum going forward, but nothing has been decided yet.
Our users would find it very difficult to find the physical space or budget dollars for even a cheap third physical machine. I second the other poster’s comment about being able to use some off-premise solution, either in a corporate data center or third-party.
We are evaluating other options for the VSA Cluster Service in 2 node ROBO deployments. I’ll share more as soon as is possible.
Hi cormac,
since the last entry is from november, i wanted to know if there are any news about the consideration of other options for 2-node ROBO deployments?
I’m afraid not. The requirement is still to have the VSACS on another host or VM at the remote site outside of the VSA cluster.