Operator
# spicedb
d
Couple questions about using the spicedb operator: - Does the operator setup HPA automatically? - Do we need to set the size of the spicedb pod or is the default enough? - What is the recommended size of the spicedb pod (if the default is not correct)
v
- No HPA - Size depends on your schema and request rate. 2 cores is always a good place to start. - don't recall the default from the top of my head.
d
For the first one, are you saying
no hpa automatically
or
dont need an hpa
? And ok, when we got the estimate from spice on how big our cluster would be in dedicated, it was 8 core 16 gb so that's what we have been using, figured it was because it needed it.
v
No HPA automatically. Given that scaling events reshapes the dispatch ring, we are still trying to figure out what's the best way to automatically scale a SpiceDB cluster. But nothing prevents you from giving it a go!
d
Ahhh ok ok, I see what you are sayng. General advice is find the size and number that works for you and stick with it.
v
yeah that's how we run the workloads right now, but we are definitely thinking about approaches to scale automatically based on metrics of the system
my experience with HPA wasn't great in the past, I've heard there is a new implementation out there, but I recall lots of pod churning, that wouldn't work super well in a busy SpiceDB cluster because the dispatch ring needs to be rearranged and the cache hit ratio will drop