Replica & Datastore Cluster

[tkutil]

Im testing Replication to a Datastore Cluster instead of a specific Datastore and it appears as thought the VMDK files are going to one datastore in the Datastore Cluster and the snapshots are going to another Datastore in the Datastore Cluster. Is this how it should work or do i have a misconfiguration issue. Also, just wondering how others are using replication for DR purposes. I was hoping I could create a Datastore Cluster on the DR side and send replicas to it versus managing datastore targets per VM and worrying about running out of space on an individual datastore.

[coolsport00]

Hi Troy -
I don't use Datastore Clusters myself, but unless you have configured your VMs to place snapshots on a different Datastore, they should go to the same storage location as the VM configuration file (.vmx). See this KB on how you can change the location where Snapshots go:
https://kb.vmware.com/s/article/1002929. Also, I didn't see anything in the vSphere Resource Management doc about snapshots being placed on a different Datastore than the VM's vmdk (https://docs.vmware.com/en/VMware-vSphe ... -guide.pdf). That's an interesting behavior.

How I have my Veeam Replication set up is I have several Datastores configured to be used by a given Replication job. That way I know where VMs are stored, and what job uses what Datastore. Since it's for DR, I don't want to keep too many restore points; in this case, for Replication, these are obviously snapshots. You're limited for Replication Restore Points anyway to 28 as VMware doesn't allow more than 32, but Veeam requires 4 for various Replication tasks. I personally only keep 14. My change rate isn't too bad, so I don't really run into space issues.

Hope that helps.

[tkutil]

Thanks Shane!
How do you decide how many vms to send to a datastore? I was just playing with my test replication job and see that it tabulates the space required as vms are added to the job so Im imagining i could use that as a tool workout the sizing.

[coolsport00]

Sure thing. Well, I don't have anything exact I use. In deciding how many VMs to have per job...again, that'll be a value different for each organization. Does the job finish in the amount of time you want it to based on x number of VMs you have in it; and, based off what transport mode you are using? Generally, you only have 2 modes you can use with Replication - nbd and hotadd. Hotadd can be pretty fast, if you have a good WAN connection. We have Gb Fibre, so I can hit GB read speeds at times. I divide my jobs up by my RPO SLAs. I tabulate how large all VMs are by job, then add an approximate amount of space for increments (change rate) based on the Restore Points I have, which is 14 in my case. Generally, my change rate isn't too great for all but about 5-8 VMs I have (spaced amongst a few Repl jobs), so my guesses are pretty good. I Replicate about 200 or so VMs, dispersed among 12 Replication jobs. Some jobs have 30-31 VMs (1 has over 40), some have 12 VMs. Generally speaking, my lowest change rate VMs, which are 'daily' Replication jobs, have the most VMs. Sorry I can't provide something more concrete, but then again...change rate is a value which will vary by organization, application, what your requirement is for Repl jobs to complete, etc. Hopefully that helps.