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

Flyte

Is there an option to configure *maximum parallelism* (for *map task*)? it is set to 5 running tasks at a time (by default I guess). I guess it is not related to the `maxParallelism: 25` I configured in the flyteadmin configuration… Correct?

So maptasks have their own parallelism controls that defaults to executing all of the subtasks at the same time. This can be controlled using the `concurrency` parameter - for example:
```    mapped_out = map_task(
        a_mappable_task_2,
        metadata=TaskMetadata(retries=1),
        concurrency=2
    )(a=a)```

Thanks you!
Do you know if there’s an option to configure it in the Helm chart (flyte-core)?
cc: <@U05794VA63Z>

<@U029U35LRDJ> We tried but it doesn’t work. Still the default parallelism for map tasks is 5. Which Flyte version do you use?

<@U059U7X5DB5>  I just tested this with the lastest Flyte version and it is working correctly. My test code:
```@task(requests=Resources(cpu='2', mem='6Gi'))
def a_mappable_task_2(a: int) -&gt; str:
    time.sleep(10)
    inc = a + 2
    stringified = str(inc)
    return stringified

@task
def coalesce(b: List[str]) -&gt; str:
    coalesced = "".join(b)
    return coalesced

@workflow
def my_map_workflow_2(a: List[int]) -&gt; str:
    mapped_out = map_task(
        a_mappable_task_2,
        metadata=TaskMetadata(retries=1),
        concurrency=2
    )(a=a)
    coalesced = coalesce(b=mapped_out)
    return coalesced```

Thank you for the answer!
This is exactly what we did. When we set the `concurrency=12`, there were 5 running pods / 7 in queue…
cc: <@U05794VA63Z>

Are all 12 Pods created in k8s? but just 7 are not running?

Ok we got the issue!
It’s the project memory quota that limits us for the maximum 5 tasks (= 5 running pods). When we set lower memory requests for the tasks - there were 10 tasks (= 10 pods) running at a time.

By the way, answering your question - it was 5 running tasks = 5 running pods, 7 tasks in queue = no pods are running. Only when pod is completed, another pod is being created. So, at a time we have 5 tasks only (based on the resources quota as I described in the comment above)

So, my question is:
Is there any metric for the *project resources* quota/utilization? That way we can understand what is blocking from our map tasks to scale…

I checked on grafana explorer and couldn’t find any metric for that, neither in here <https://docs.flyte.org/en/latest/deployment/configuration/monitoring.html>