MSA1000 With two Controllers and two embbebed Swtiches

[Rafael Gersol Ayllón]

Hi.

We have some doubts about tha High Availability.

We've recently adquired a a High Availability Kit for a MSA 1000.
Now we have a MSA1000 with two controllers and two embedded 2/8 2Gb Switches.
We have installed following the instructions And now the REdundant LED is active and in LCD display a "Redundant OK" message is showed.
we have to say you than the two clusters haven't dual HBA, It's to say, each node has a HBA only one.
In this case, What type of High Availability can we to have?

We known than:

1. Switch 1 is connected with controller 1 and switch 2 to the controller 2, and Switch 1 can't work with controller 2 and vice versa.

2. If ours nodes haven't dual port, we can't use two paths,and if controller 1 hangs, nodes connected to the swtich 1, don't will see the controller 2, and we have to connect manually the nodes to the swtich 2.

Are these afirmations correct?

How can we use the redundancy in controllers and switches? or
How can we do a test of availability?

Any configuration special is required for enable Redundancy?

Thanks for your attention

[Jefferson Humber]

Yes, you are correct. If you loose one Controller/Switch you will loose one node of your cluster due to single paths.

The redundancy in your situation is you won't loose both halved of the cluster. Ideally though you need 2 HBA's per node to get the max benefit out of your configuration.

Jeff

[Steven Clementi]

Without dual fibre channel hba's, you can not have any redundancy. There IS NO redundancy since you are not connected to the "redundant" controller.

Statement 1: Yes and No. Technically, you can connect both switches with an ISL(which would allow Controller 2 to talk on Switch 1), but you would have to configure your zoning so that your HBA can ONLY see one MSA Controller. Dual Pathing on a single controller is NOT supported.

Statement 2: Correct, for the reason stated above. I would NOT manually move the connections.

The only way to get the redundancy is to get a second HBA. UNtil that time, it is better to just remove the second Controller since you really do not want the controllers to fail over if there was not real failure.

Again, to enable redundancy... you NEED 2 HBA's or a Dual Port HBA.

Steven

[Steven Clementi]

to enable "redundancy" on your storage, that is.. not your cluster. A cluster, usually by default is a "redundant configuration". HOw redundant depends on all of your other components.

You cluster may have some issues where the services hang for a bit while the controllers fail over, but technically... you should be back up and running within minutes, if not sooner.


Steven

[Rafael Gersol Ayllón]

Thanks to all for your time.

Then, thinking about your replays. I want to launh a question:

If you have a single controller with a single switch and any of the two elements hangs, my two clusters hangs too. But if I have other set of switch and controller, Could I use them? Although the connections are made manually.

It's better have two sets of controllers and switches than an only one set, no?

In resume, My question is , How I can to take of adventage of these elements while we have an Only one HBA per node?

Our Idea was to have redundancy in controllers and swtiches and connect all nodes in a switch and controller. If a fail occur, to use the other switch and controller, connecting the nodes in its.

What you think about this?

Thanks.

[Steven Clementi]

"But if I have other set of switch and controller, Could I use them? Although the connections are made manually."

The MSA1000/1500 does not support multiple connections on the same HBA. Yes, it would probably work since there is still only one path, but you potentially might confuse the controllers and can potentially have problems.

"It's better have two sets of controllers and switches than an only one set, no?"

Having 2 sets is irrelevant. Each set does not interact with the other in any way. Now, depending on how you configure the second set (potentially as a software based replication solution)... it may a more "redundant" solution.

"In resume, My question is , How I can to take of advantage of these elements while we have an Only one HBA per node?"

The simple fact is HP does not support a single HBA with redundant controllers. If you are short on PCI slots, then get a dual port HBA. IF you are short on money, then there is not much else you can do except try to find the hba's cheaper elsewhere.

"Our Idea was to have redundancy in controllers and switches and connect all nodes in a switch and controller. If a fail occur, to use the other switch and controller, connecting the nodes in its.

What you think about this?"

I think you are asking for trouble, but the config would probably work...unsupported.. but work.


Steven

[Rafael Gersol Ayllón]

Thanks for your atention.

I will negotiate Acquisition of several HBA's in a short period of time. We will then use dual Path's with dual HBA's per node. While we use redundancy only to replace parts, when these fails.

Thanks

[Rafael Gersol Ayllón]

Thanks