vSAN stretched cluster - DRS

[Sharantyr3]

Hi there,

I'm currently in the configuration stage for a 4+4+1 stretched cluster in all flash 6.7 u1

I got a few questions about DRS and couldn't find clear answers or outdated ones.

For instance, "official" guide (Per-Site Policy Rule Considerations) states "It is entirely possible to have a VM Storage Policy using the Affinity rule placing data in the one site, with a VM/Host Group Rule placing the VM in the alternate site."

The provided screenshot after this sentence show VM components "active" on a specific site.

In my configuration, VM components are "active" on all ESXi on both sites (using a PFTT=1 and SFFT=1 with erasure coding).

Based on recent papers about "read locality", vSAN is supposed to read on the site where the VM is running. And also I can't choose which site is supposed to be prefered for a specific VM's data, on storage policy, because I have PFTT=1.

So the sentence quoted above seems incorrect to me with PFTT=1, what don't I understand ?

Also, is it still a requirement to do host groups and VMs groups based on site when you do PFTT=1 ? (Apart from separating VMs for site disaster resilience purposes)

How do you manage these groups ?

I don't plan on asking other IT admins to populate DRS groups as they will forget, so I was thinking to use TAGs + powercli to populate DRS groups, but I'm looking for "Here, we do like this..." infos.

[peetz]

Greetings,

in a vSAN Storage Policy the "Affinity Rule" can only be used with PFTT=0. In this case the VM will be stored on only one site, and you decide on which by setting the Affinity to either the primary or the secondary site.

Only in this case you should make sure that the VM can only run on hosts that are on the site where it is also stored (by using VM/Host DRS rules) to avoid inter-site traffic.

If you use Storage Policies with PFTT=1 for all your VMs then you don't need these rules, because the read locality feature will make sure that data is read locally.

Andreas