HTTP auto scaling practicality

[rreeves8]

I came across KEDA researching ways to save cost on our pods by scaling to 0. And I’m surprised to see the large number of auto scalers offered by default, non of which include http events? Why is this? Is there some caveat that makes it impractical to scale solely based on http events? Why is this not a more widely used system?

Im just curious as I would’ve thought http events would be the main scaler provided by KEDA. However it seems it’s not offered by default.

[linkvt]

Not sure if still relevant but HTTP scaling from 0 is a bit more complicated as you would need to buffer requests (-> this is what the HTTP Addon does) while starting the pod in the background (-> this is the normal scaling KEDA does).
This is complex for larger environments as e.g. all requests would be buffered in memory, etc and keda would be in the request path. Usually keda just observes the "request paths" in form of a queue size, something in a database, ...

Closing due to inactivity. If this is still relevant, please comment here or open a new discussion and we'll reopen or address it.
If it's a bug, please file an issue instead.
Thanks!