HA and Single Points of Failure

[FlVEBYFlVE]

Hello,

I currently have three 6.7 ESXi hosts.  Two of which are using local storage and one is using ISCSI storage.  None of which are currently connected to vCenter.  All three are close to max on resources as well as the ISCSI storage.

I have a request to create HA redundancy for these three systems.

Best practice suggestions for this are welcome.  However, my primary question is:  When using shared storage for HA what are people doing to mitigate the risk of the shared storage being the single point of failure?  How can you have true HA with shared storage?  Assuming you put the necessary hardware redundancies into place, that device is still a single point of failure for your HA.

Is there a way to create HA for local storage?  Like a using witness server that could synchronize between active and passive nodes?

Thank you in advance.

[daphnissov]

The most common way of providing for high availability of virtual machines is to use a replication solution to replicate VMs from a source (shared) datastore to a target (also shared) datastore. This is commonly at a secondary site. Backups would be used intra site to provide for additional availability. But all these solutions require a management solution like vCenter, so you have that problem. And, even before getting there, you're already resource constrained in using local storage. So your bigger problems need to be solved by hardware before even getting to the HA question that can be solved by software. It sounds like the order of priorities should be shifted with your management.

[FlVEBYFlVE]

Thank you for the response.  Hardware is not an issue.  We can purchase whatever hardware will be needed to accomplish this as well as any licencing, including vCenter.