After discussion with Jay:
- About the naming, we can stick to the term of "mixed" availability
(rather than "aggregated" or whatever) to keep continuity with previous
products
- The endpoint would accept query by ids or tags
- Resulting series would provide datapoints with a key/value map of ratio
per state. So states such as "unknown" and "admin" are not discarded.
More
interestingly it would also be usable with string types
On Tue, Aug 30, 2016 at 3:51 PM, Joel Takvorian <jtakvori(a)redhat.com> wrote:
On Tue, Aug 30, 2016 at 3:11 PM, John Sanda <jsanda(a)redhat.com>
wrote:
> I like the idea of aggregated availabilities, but I don’t know that it
> can easily be simplified to up/down. Let’s say we have 3 Cassandra nodes
> deployed with replication_factor = 1. If one node goes down we are at 66%
> availability.
>
You're right, it would be even better to keep a "up ratio" information for
each point. And it will still be easily reduced to a binary "all UP" or
"any UP" by the client if desired.