05-12-2013 01:06 PM
We are planning to upgrade our current fabric set up consisting of 6*48000 directors per fabric (ruuning in half mesh topology) to DCX in Core-Edge topolgy(2 cores and 4 edges).
We were planning to perform like this
1. Build the new fabric using DCX in Core-edge topology.
2. Identify the systems sharing same storage ports,
-move server HBA's from current 48000 to DCX edge switches in Fabric 1 alone
-move storage port from current 48000 to DCX core switches in fabric 1 alone.
3.Once these new paths through DCX are up on servers, then perform the same for Fabric 2.
Please let me know, if there are any better thoughts than this, the idea is to migrate without any or little impact to production.
05-13-2013 06:25 AM
in your scenario you will move physically the Devices in one Fabric at one time, so lets say in one night.
In this case you have a risk that there are some upcomming errors and you have a longer outage of one fabric.
Another scenartio is, that you bring up both DCX-Fabrics. Now connect 48000 FabricA via FCR to the new DCX FabricA.
Now you can move step by step each Host and each storage Device. You can check individual if this is working.
This takes a longer time, but the risk of an outage is only by one Host or one Storage system.
When you finnished FabricA, do the same with FabircB
05-13-2013 07:53 AM
when he connect old 48K via FCR to DCX is has the following advatages:
Both Fabrics will not merge.
He can move each host and storage seperatly. Connection can be done via LSAN-Zones for a short period.
So you can move one host, check everything and go on with next host or Storage.
Because he asks about other possibilities to to the job i post the way with the FCR. And this has on my point of view the lowest impact.
05-21-2013 05:13 AM
I am planning a migration from 48000s to DCXs also. I am interested in finding out more about using FCR to minimize impacts. I use a IBM San Volume Controller to virtualize all my backend storage and present luns to my hosts via the SVC. Could you provide some more info or point in a direction where I could get a better understanding about how using FCR would work.
05-21-2013 06:24 AM
That sounds like a good idea, I am yet to read any docs related to migration using FCR
I would like to get clarification on the below queries,
When we physically moving the systems from old switch(48000) to new switch(DCX), do we not experience link going offline while the cable is plugged out from old switch and plugged to the new switch. Or does the LSAN zoning ensures that the the virtual server and storage are able to talk to each other and at the server end, this do not go offline. Please let us know your take on thi
05-21-2013 07:42 AM
when you disconnect a cable, the link will go down. That is Layer1.
LSAN-Zoning ensures that devices which are members of different fabrics can talk to each other.
So what you can do:
1.) All Hosts and Targets are connected to 48000-Switch. -> no impact, no outage.
2.) Now build a FCR Topoligy between DCX and 48000-Switch. -> no impact, no outage.
3.) Build up LSAN-Zones for i.e. the first Host. -> no impact, no outage.
4.) Disconnect the first Host from 48000 and reconnect to DCX. -> no communication while changing the switch for the first Host, outage for the first host only.
Then you have to move step-by-step all hosts and targets and insert the new "normal" Zoning and the remove the LSAN-Zoning.
05-27-2013 07:05 AM
Thanks for clarification,
Also when we do it for one host at a time for single fabric, with the multipath the server could still run production on the redundant paths on second fabric. I am looking at the pricing for running this migration. Coz we have a complicated half mesh topology (6 switches per frabic), we might require multiple FCR's if the host and storage are not running on the same switch.
09-11-2015 01:23 PM
Can you please help me with a step by step instructions to do FCR
We are planning 2 Fabrics migration from 48K to DCX-8 with 14 edge switches.
And another 2 Fabrics migration from 48K to DCX-4