We have a setup running metaflow, we use UI sparingly and we noticed consistent under utilization of ui-backend task. It requests 2vCPU and 16GB of RAM, while in our case we could get away with running it with the same resources as the static container.
It would be great if resource requests for each task were exposed as variables in the main module. Our cost difference would be a whooping $80 monthly (close to nothing compared to the compute resources), but still it would be nice to be able to provide a correct size and optimize base costs of metaflow being deployed.
something like:
variable "ui_backend_cpu" { default 2048 }
variable "ui_backend_memory" { default 16384 }
(and counterparts for other tasks) would be nice.
We have a setup running metaflow, we use UI sparingly and we noticed consistent under utilization of ui-backend task. It requests 2vCPU and 16GB of RAM, while in our case we could get away with running it with the same resources as the static container.
It would be great if resource requests for each task were exposed as variables in the main module. Our cost difference would be a whooping $80 monthly (close to nothing compared to the compute resources), but still it would be nice to be able to provide a correct size and optimize base costs of metaflow being deployed.
something like:
(and counterparts for other tasks) would be nice.