Start a Conversation

Unsolved

This post is more than 5 years old

I

398

December 5th, 2006 08:00

Redundancy Overkill?

Newbie here so please be gentle. :p

We currently have a 2-node active/passive server cluster connected to 2 fabric switches. Since I am bad at explaining things, I will try and explain our setup as follows:

- Node1 (active) has two HBAs installed (hba1a and hba1b)
- Node2 (passive) has two HBAs installed (hba2a and hba2b)

hba1a and hba2a are connected to fabric switch #1 and hba1b and hba2b are connected to fabric switch #2. The CX300 is also connected to the fabric switches, but this point is moot in the questions I am about to ask.

Wouldn't it be alright to have only one HBA installed in each node having Node1 connected to fabric switch #1 and Node2 connected to fabric switch #2? So for example, if fabric switch #1 went down, then Node2 would become the active node and vice-versa??

I know that you can never go overboard with redundancy, but seeing that our fabric switches are only 8-ports and we have a requirement to add additional clustered servers, I am trying to free up a couple of ports on the switches without losing redudancy altogether.

3 Posts

December 5th, 2006 13:00

The answer depends on your tolerance for component failure.

You can:
- connect Node1 to Switch1 via HBA1a and have it zoned to SPa0 and SPb0
- connect Node2 to Switch2 via HBA2a and have it zoned to SPa1 and SPb1

The caveat here is that a single hba, cable, or switch failure will cause a cluster failover event to occur.

If you were to crossconnect everything...

- connect Node1 to Switch1 via HBA1a and have it zoned to SPa0 and SPb0
- connect Node1 to Switch2 via HBA1b and have it zoned to SPa1 and SPb1
- connect Node2 to Switch1 via HBA2a and have it zoned to SPa0 and SPb0
- connect Node2 to Switch2 via HBA2b and have it zoned to SPa1 and SPb1

... then only a true node failure would invoke the cluster failover. All paths would be protected.
No Events found!

Top