vSAN Streched cluster, drs placement with no site affinity?

Hi folks,

I’m looking for a streched cluster recommendation, we have now a vSAN cluster and this year we want to extent this server room with another one in other building inside the factory ground to improve availability, All the networking and facilities work is in progress and we have are goind to start the server’s procuremenet for the new server room soon.

But I have one question regarding DRS placement if some vms dont’t have any site affinity with SFTT =1 and PFTT =1 .

Use of DRS host/vm affinity rules is mandatory with streched cluster?. My understanding is that if you set site affinity in vSAN then you have to create the rule, but if site affinity is not set the DRS will do the initial placement on the best site and host and the VM could even change from one site to another is there is performance improvement without impact and with read locality in the compute host site. Is my understanding right or it will not work as expected?

Also, for those VM with vSAN site affinity, is DRS aware of it so it will place the VM on the site or we should create de Host/VM rule to place it correctly?

Just to give more information, we plan to use 3 vSAN Policies

– PFTT 1 SFTT1 no site affinity for production VMs with no GuestOS or Application clustering

– PFTT0 SFTT1 with site affinity and DRS Host/VM affinity for SQL AlwaysOn or WSFC or other kind of application level failover. one vm on each site

– PFTT0 SFTT1 no site affinity for non production/dev VM

View Reddit by desertspotterView Source

Related Articles

One Comment

  1. To be honest I would create site affinity and DRS rules for every workload.
    This way you can make sure that you do not get an unbalanced cluster, depending on the number, size and performance needs of the workloads without actual site affinity.

    Moreover this reduces traffic on your links between the datacenter.
    Sending storage traffic from VM1 in DC1 over to the vSAN cluster in DC2 just because someone moved it over there, or DRS moved it (because it had instruction to do otherwise) may impact your prod workloads at some point.

Leave a Reply

Your email address will not be published. Required fields are marked *