On Fri, Apr 29, 2022, 10:17 AM Gk Gk <ygk.kmr@gmail.com> wrote:

Hi All,

I need information about availability zones in nova. I tried googling but cant find enough information. My questions are ,

1. Why is it that we have two concepts of aggregates and AZs ? Is one not enough ? Like exposing aggregates and creating flavors with extra specs to match ? Why do we need AZs also ?

AZ isn't for picking the right type of host. It's for picking hosts that are distributed. Maybe you want a host with local disk. Great, you use aggregates / flavor specs for that. But what if you want instances for a myself cluster or setting to each be in different racks? Then you need AZs.

2. Why is it that one node should only be a part of one AZ but not two ? whereas in the case of aggregates, it can overlap ?

Overlapping failure domains would be pretty confusing and pointless.

3. Also why cant we expose only aggregates like AZs but block the compute member list to the users ? Doing this way will serve the purpose of AZ as well ? Why we dont want to expose aggregates as AZs ?

They're mutually exclusive concepts.

Thanks Kumar