High Availability, would something like this work ?

[AUS_Mike]

Don't know why but an idea came to me on how a H/A solution may work at least in my case,
Without Visio its a little difficult to draw so bare with me

"Solution from top to bottom"

=====CONT1=========DAS - MD1000=======CONT2=======

====Server 1 NodeA ===== Server 2 NodeB ===== (Single extSAS connection to each controller)

======Node A iSCSI ====== =======Node B iSCSI ======

Server 3, Application

The only issue I could see, is if a server failed during writes which could cause some issues.

Would something like this work ? storage isn't my forte and I know I have generalised this with simplicity.

 

[jgreco]

In general, the lower end high availability solutions tend to assume that a head will fail but that the underlying storage remains functional. In such a case, a redundant head may be part of the solution. I think that's what you're trying to show.

Active/active storage heads on a single storage array are somewhat tricky to implement because the thing that makes a storage system fly is caching; ZFS lives and dies by its caching, and it is not possible to implement an active/active head solution with ZFS. A simple cluster-aware filesystem might simply feed requests through directly to the storage, eschewing all the caching, and performing sync writes, which is more or less what VMware's VMFS does.

You can implement a dual head active/passive setup with ZFS using a heartbeat strategy, which is how systems like Nexenta implement HA on ZFS. You can only implement active/active "sorta" by having two pools, one head acting as active head on pool 1 and passive head on pool 2, and the other head acting as passive head on pool 1 and active head on pool 2. There's still a bit of a mess when a cutover has to happen, because the backup head has to reliably detect that the primary head has failed, and then forcibly import the pool.

 

[AUS_Mike]

only looking for an active / passive solution, and yes you are correct, having 2 x ZFS controller heads,