Linking Autostart Clusters with replistor

Thanks.

Once VQE is assigned time to work on this, we'll set it up here.

The site is a -massive- airport, so we want to get it right the first time.

Al Verrier

Software House Technical Support

781-768-0235

70 Westview St.

Lexington, Ma. 02421

This e-mail may contain or have attached to it information or technology the export or re-export of which is restricted by U.S. laws and regulations. By viewing this e-mail and any attachments, the recipient agrees to the following: the recipient's use of the contents of this e-mail and any attachments shall comply with all applicable laws, rules and regulations, including, without limitation, U.S. international trade laws and regulations, and (ii) the recipient may not transfer or otherwise export or re-export any information or technology contained in this e-mail or any attachment except in full compliance with the export controls administered by the U.S. and other countries and any applicable import and use restrictions

Hi Joel

Good to "talk" to you again.

My concerns were along the lines of your email. If using a production and a recovery domain model, I was concerned about things like replication pointed at the 'inactive' node in the recovery domain. Using the DR alias would avoid that situation quite nicely.

We would use the 4 license autostart/replistor model, and forwarding to the alias would simplify things.

Our normal configuration with the three node design is for the 'failback' to be manual.

If we proceed with the 2RG model can you think of any other potential pitfalls? We are looking at using a composite datasource on the two clusters.

We're going to qualify this design in house, any pointers would be most welcome.

Happy holidays.

Al Verrier

Software House Technical Support

781-768-0235

70 Westview St.

Lexington, Ma. 02421

This e-mail may contain or have attached to it information or technology the export or re-export of which is restricted by U.S. laws and regulations. By viewing this e-mail and any attachments, the recipient agrees to the following: the recipient's use of the contents of this e-mail and any attachments shall comply with all applicable laws, rules and regulations, including, without limitation, U.S. international trade laws and regulations, and (ii) the recipient may not transfer or otherwise export or re-export any information or technology contained in this e-mail or any attachment except in full compliance with the export controls administered by the U.S. and other countries and any applicable import and use restrictions

Al,

I think the 2 RG method is the best because it simplies DR failover. That being said, you'll really want 3 RGs, one for Production Services, one for DR services and 1 for DR Data Replication Reception. I can't think of any general pitfalls with this configuration but can certainly assist if anything comes up.

Happy Holidays to you too.

Al,

I hope all is well with you, it's good to see you utilizing the forums.

Using RepliStor to transfer files between clusters is certainly an option but has some technical challenges which you no doubt have been pondering.

I personally like using the same configuration as the 3-node cluster which you describe below. The only difference is that the 4th node is mirrored from the 3rd node. In that configuration, you may have two 2-node clusters or a single 4-node cluster. The 4-node cluster complicates the configuration of the resource groups since the local nodes and DR nodes utilize different data sources. Alternatively, you may have 2 Resource Groups.

Another option which is technically more robust but offers many more management headaches is to have RepliStor configured to send to a node alias on the DR side. With this configuration, it's best to think of the two clusters as independent, a production cluster and a DR cluster. Again, these can technically be part of a single 4 node cluster but the 2 roles are still present, production and DR data recipient. The DR node hosts a mirrored data source and a receiving alias. RepliStor is configured on both sources to send data to the DR alias rather than to a specific DR node as would be done above. From a license perspective, 4 RepliStor licenses would be required in this configuration whereas only 3 would be required above.

In either case, it's best practice to integrate RepliStor on the production side with AutoStart. The failback procedure and use of RepliStor for recovery should be a manual process and is generally not recommended to be placed within a Resource Group.

I know the above is quite wordy and complex so please let me know which parts I can clarify.

Cheers!