Replacing old RP environment

Cool.

I just found out that the old clusters are 4.4, same as my new one.

It's my lucky day.

Once I get the data off of the 2 old RP Clusters - is it easy to decommision them (make them go AWAY in the GUI)?

thanks

yeah - if you are connecting them to do concurrent replication, once you are done with the clusters being connected, you can disconnect them in the CLI as the boxmgmt user and then they will no longer show in the GUI topology.

A source LUN cannot be replicated by two different RP clusters. It can be replicated to different RP clusters as multiple targets, i.e. fan-out.

Regards,

Rich

So I guess my next steps are:

Remove licenses from Lower set

Disconnect them.

Prep them

Add to original (upper cluster pair)

Re-add licenses (since more XtremIO need to be added)

Source XtremIO-A --->RP---------------->RP-->DR XtremIO_B

         |                                                          

         |

        ^                         new RP

New XtremIO_C -->RP---------------->RP--> New DR XtremIO_D

In your example I am assuming you have 4 RP Clusters (A cluster is defined as a set of RPA's at a single site). It looks like each XtremIO array is part of its own RP cluster - is that correct?

Sounds like you have 2 clusters (original) with 6 RPA's each and 2 clusters (new) with 8 RPA's each. You can't remove RPA's from a cluster online like that and the "disconnect" is only for connecting clusters to each other over the WAN and not actual RPA's in a cluster. You can connect the 8-node RPA cluster that houses XtremIO-C to the 6-node RPA cluster that houses XtremIO-A, but you will only use the first 6 of the new 8-node cluster as target RPAs since they work in peer groups.

For your steps, it could be simpler. If all of these clusters are up and running right now (again, assuming 4 RP clusters), then all you would like need to do is set another WAN gateway for XtremIO-A cluster to talk to XtremIO-C cluster over the WAN, use the Deployment Manager to prepare and connect those clusters, and then you can reconfigure the XtremIO-A cluster's CG to add additional copies to their RSETs that consist of target LUNs/Jvols from the newly added XtremIO-C cluster. This would assume that the clusters are already licensed for their respective arrays and that will carry over when you connect them.

Let me know if that helps or if I am not interpreting your topology correctly with 4 clusters.

Another question - original clusters have 6 RPAs.

The new clusters have 8.

Can I simply "disconnect from cluster" 2 of them to match original before connecting?

thanks

ahh good catch Rich...I think I thought they were fresh and clean....could you just disconnect them from each other in that existing RP system or does it have to be a reformat of the repo to break them apart?

Nice to know I don't have to make RPA # per cluster match.

The second pair of RPA clusters (8 per cluster) are NOT in production yet. No CGs, just a Repo

I would think I need to disconnect them from each other and add them BOTH to Cluster "A".

I believe (finding out in 10 minutes) that all the clusters use a 192.168.200.x addressing scheme.

I was able to connect cluster C and Cluster D without specifying a WAN gateway.

I would perform a disconnection, remove any config, remove any licensing and reformat the repo.

You will need to performing a reformatting procedure in the old RP system before adding any of the clusters into the new RP system if that is your intention. You cannot just add a cluster to another system when it has been part of a previous RP system.

ooh ok...you want the LUN in the XtremIO CLuster-A to be replicated to BOTH the XtremIO-C and XtremIO-D RPA clusters as targets?

I only interpreted as you wanting to replicate it to Cluster-C from the arrows since you can't have cascaded replication from A-to-C-to-D. So if you want them both to be targets, then you would connect them both (C and D) to A using the DM/CLI twice to do them both.

Yeah - you might not need a WAN gateway to do it if they are all on the same flat network, I just throw that out there because that is usually where 99% of the issues are when connecting clusters from my experience

Ahh, OK...then I stand corrected Richard Coccoli .....Do what Rich states....he knows best!

Thanks Rich

Ill be DOING it in about 1/2 hr.

When I reformat the repo do I have to detach each node from the cluster ?

reformat from first RPA then "use existing repo" from others?

thx