Open malomarrec opened 2 years ago
How urgent is this? It's definitely on my radar but didn't get a lot of attention yet and it needs to be figured out what that means for fairness and executors would need to be able to identify their max capacity. I'd say a rough estimate here would be 5d, with some questions to answer before we go into implementation.
Problem
Customers running batch changes server-side do so for a wide variety of workloads: from running simple utilities like sed, to running complex refactoring tools. Some of those tasks require a small amount of resources (CPU, RAM), but some require lots of CPU/RAM (eg. running an AST-refactoring tool that requires the running the JVM).
Today, we have a one-size fits all model for server-side jobs, that is set by the site admin:
This model doesn't accommodate the heterogenous workloads required.
Solution
It's likely that we'll allow users to select among a small set of instances to use for jobs, with opinionated defaults.
Pricing
See thread