HPE EVA Storage
1844561 Members
3228 Online
110233 Solutions
New Discussion

Re: Load balancing on Continuous Access

 
SOLVED
Go to solution
Ian Dennison_1
Honored Contributor

Load balancing on Continuous Access

I have a 2 Fabric SAN, each Fabric connects via a pair of MP400s to a remote SAN, with an EVA at each site.

One thing I noticed when setting up was that the first Fabric configured did not allow the Management Appliance at the local site to see the remote EVA. It was only when I configured the second Fabric that this was possible.
Now I am replicating across the link, traffic is only going across the second Fabric.

Question: Should the CA product auto load-balance traffic across Fabrics? Is there an option I missed on the setup (Both were configured using the same commandsets)?

Should I disable the second fabric and see if the first gets used?

Share and Enjoy! Ian

PS. The Fabric OS documentation for the MP400s is a steaming pile of guano; wrong Fabric OS, wrong commands, wrong dependencies. I will sanitise my own instructions and upload them later.
Building a dumber user
3 REPLIES 3
Ian Dennison_1
Honored Contributor

Re: Load balancing on Continuous Access

Clarification: It is the MP Router XPath 7.4 manual which is irrelevant, however it is the only document I found that contains the combination of both the theory planning and implementation steps (albiet with wrong commands) to implement FCIP Tunnels.

Share and Enjoy indeed! Ian
Building a dumber user
Peter Mattei
Honored Contributor
Solution

Re: Load balancing on Continuous Access

CA EVA uses a very efficient replication protocol that only takes one roundtrip per data exchange.

In order to ensure consistency traffic for a single DR group can only use one path. This path is called a tunnnel.
So, if you create just one DR group you can only replicate through one tunnel and thus one fabric.
However in case of a failure the EVAs can rebuild the tunnel over any controller and fabric!

As you know all Vdisks in a DR group are owned by one controller.
So for an example EVA1-CtrlA owns all Vdisks of the source DR group while EVA2-CtrlA or CtrlB may own the Vdisks of the destination DR group.

This also means that if you create 4+ DR groups you can have 4 tunnels and utilizing all 4 controllers and both fabrics.

See my attached drawing where I show the relationship between DR groups and tunnels.

Cheers
Peter
I love storage
Ian Dennison_1
Honored Contributor

Re: Load balancing on Continuous Access

Peter,

Thank you for the input. I have already set up "tunnels" for the FCIP connection between routers, so are you describing another "tunnel" that exists at a higher level (or Layer for the CCNAs in the audience) that is controlled by the EVAs?

If so, how do I control (directly or indirectly) which controller is used?

Share and Enjoy! Ian

Attached is an instruction document I created for establishing FCIP Tunnels which may be useful. I did this on a clean-site so use with care on existing sites. All WWNs and Device Names anonymised.
Building a dumber user