Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

HI team, QQ, for flyte map tasks, we have can use resources to specify the requests and limits, is it also possible to pass in the node pool to run this task on?

```map_task_partition_outputs = flytekit.map_task(process_map_task_partition, concurrency=300)(
        map_task_partition=map_task_partitions).with_overrides(
            limits=flytekit.Resources(cpu="4", mem="42Gi"),
            requests=flytekit.Resources(cpu="4", mem="42Gi"),
            retries=3,
    )```
Example,== ^^^ is one of our map task function call.

hey varsha - just wanted more information, could you elaborate on why please?

there’s no way to set the node pool currently

We are partitioning our eks node pool across different instance types/team etc

So when we run Map task for a specific project, we would like to provide a node pool to the task to let it know where to run

would it be possible to achieve this with taints and tolerations?  <@U029U35LRDJ> do you remember if it’s possible to add tolerations by task type?

adding the concept of node pools explicitly is a bit tricky.  flyte aims to be k8s native, but we still try to abstract the internals of k8s away from the user when possible and reasonable.  also node pools aren’t a k8s construct right?  they’re eks/gke constructs

you can apply labels/annotations to a given project or project-domain <https://github.com/flyteorg/flyteidl/blob/master/protos/flyteidl/admin/matchable_resource.proto#L116>

but i’m not sure how those tie into node pools

you could use taints and tolerations maybe too