Building a HighAvailability Shared Storage Cluster

[Squazzer]

Hi there

I’m looking to build my own HighAvailability Shared Storage Cluster.

The ideal setup would be something I can build with my own hardware, and with as little hardware as possible.

In my current setup I’m running two Dell Poweredges connected to a Dell PowerVault DAS filled up with 2.5" SAS SSDs. Giving each Poweredge direct access to each SSD in the DAS enclosure.

Right now I’m using a combination of Windows Storage Spaces and Active Directory to handle the HA setup. But I would like to try something else.

I’ve been looking at TruenNAS Scale. And even though their website tells me that TrueNAS Scale should be able to make a HA setup with a DAS & 2 controllers, I’m being told (by iXsystems) that it’s only possible to do with their own hardware, not with my own.
But their description of a Shared Storage Cluster matches very well what I’m trying to build; “Two nodes are used to manage a shared ZFS pool.”

Is this possible to build a HA Shared Storage Cluster with my own hardware, but using some kind of Linux?

[risk]

It sounds like you’re trying to build what some folks in the industry call “medium availability” setup where you have the drives accessible through either one of the two hosts but not either? And switching between the two requires some process where your data is temporarily not available.

There’s a thing called keepalived, there’s also a thing called drbd if you want to mirror your storage across nodes.

---

If you really want high availability, you need more than 2 nodes so you don’t end up with split-brain.

But, maybe you could do ceph pretending each drive is its own independent failure domain (even when they’re on the same host?)

[MadMatt]

Sure, have a good look at this:

Home