Hello everyone. The company where I work is going to migrate from Hyper-V to VMware vSphere Enterprise Plus, so after almost 6 years using Hyper-V, this is a totally new world for me. In order to do that, we've acquired new hosts that will be spread between two sites, one is the main site and the other is the site for DR. We've also acquired two IBM FlashSystem FS5035 storage arrays. One will be placed in the main site and the another will be in the remote site. The sites are 40 km (~25 miles) apart from each other and are connected with a direct 8 Gbps fiber link.
I've read in IBM documentation that these storage arrays support synchronous replication through a feature called HyperSwap. It seams that I've to install a DR orchestrator in one site (IBM Spectrum Virtualize 3-Site Orchestrator) and it will monitor and control the replication process. I also have to install a quorum application (that will be on a separate building on the main site) to break the tie and determine what system must interrupt I/O operations.
1 - My big doubt is what is the best approach to deploy VMware vSphere in this evironment, where the replication is orchestred on the storage level?
2 - How would we manage a failover to the other site? Can I do an automatic faiolver with this architecture? How will the VMs behave in the second site during the normal operation and a fail over since it is a storage level replication?
3 - Since the DR site is a copy of the first site, is a good practice to create a single unique cluster with the nodes on both sides, since the storage layer is using synchronous replication?
We wanted to use VMware SRM, but at the same cost of licensing it we would be able to add two hosts in this architecture, so a loud NO from the financial director.
4 - I also have doubts regarding the migration procedure. We have multiple relational databases (MS SQL Server, Oracle, PostgreSQL, MySQL). All together must have almost 40 TB of data. This must be a hot migration and our DBA suggested adding VMs deployed in the new infrastructure to the existent database clusters and after the migration, keeping only the replicated database and then eliminating the original replica. I'm concerned about this operation of adding and removing VMs in critical clusters and also with licensing implications. How do you migrate big database workloads that require a hot swap between the older and new infrastructure? Since we're currently using Arcserve UDP for our backups, we'll use it to migrate our non "real-time change" workload (I've heard of VMware VM Converter, but we are running Linux distros that aren't in the compatibility matrix and we don't want run on troubles and surprises) and then they offered for us a product called Arcserve Migration. Have you even used it or do you have suggestions on better tools for this kind of operation?
Thanks for your attention and patience for reading this big text. I can't wait to see the answers!