topic Controller Failure in HPE 3PAR StoreServ Storage

Controller Failure

kunjuttan — Wed, 14 Dec 2011 07:16:10 GMT

Hi All,

I am having 3 hp storage(EVA8K,EVA4K and EVA3K).I have lot of Vdisks created in all these storages.Some Vsiks preferred path is Cntlr-A, while some other vdisks are controlled by Cntrl-B.Now my question is if any of this controller fails, what will happen to the LUNs owned by that controller. Whether the other controller take over these vdisks from the failed controller or not?

Re: Controller Failure

Torsten. — Wed, 14 Dec 2011 07:21:22 GMT

The other controller will take the ownership and will present the LUNs.

Re: Controller Failure

kunjuttan — Wed, 14 Dec 2011 07:26:50 GMT

Thanx Torsten for the quick reply.

So if I have to do a controler reboot there is no need to change the prefered path to other controller,isn't it?

Is there any special configuration or anything else required for this?

Then what is the role of HA cluster in Storage,which has become a common term in storage as well.

Re: Controller Failure

Torsten. — Wed, 14 Dec 2011 07:39:15 GMT

The user guides describes the options (failover only, failover/failback, etc ...) very well. The setting depends on your needs. Not every setting is supported for every OS - refer to the manual.

>> Then what is the role of HA cluster in Storage...

Depending on the array model, the nodes work like a cluster (similar to servers), e.g. on the new 3PAR arrays (up to 8 nodes inside the array).

Re: Controller Failure

kunjuttan — Wed, 14 Dec 2011 07:48:53 GMT

HI Torsten,

Normally I used to give as no preference option for Vdisks. Is that ok for automatic fail-over or do i have to give anything else like Path A-Failover,Path A-Failover/failback etc.

And regarding cluster, we are achieveing the same functionality of failover/failback with this preference method.then what is the relevance of cluster configuration?

Re: Controller Failure

Torsten. — Wed, 14 Dec 2011 07:55:49 GMT

As said, refer to the user guide or online help for these settings.

In general, you can decide which controller owns a LUN and what should happen after a failed controller is online again (failback).

In case of a controller failure the "All LUNs are brought online to the
surviving controller." (from user guide).

The understanding of "cluster" may depend on the point of view.

In general, the controllers of most arrays work like a cluster inside the array, but usually you can even "cluster" certain arrays - but this depends on the model and options.