Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.

Prefect Community

Hey everyone, we are using Prefect with k8s. Our flow runs are always marked as "Failed" at the beginning and only later as "Running" and "Completed". I guess pods with flow runs don't start fast enough so Prefect marks them as "Failed". Is it possible to somehow increase this time limit? We get unnecessary notifications about failed flow runs. :confused:

I read a comment on this somewhere. But you could set appropriate time diff in one of the following settings
```PREFECT_AGENT_PREFETCH_SECONDS='15' (from defaults)
PREFECT_WORKER_PREFETCH_SECONDS='10.0' (from defaults)```


Not entirely sure where I read it though.

alternatively while running agent / worker from cli `--prefetch-seconds` is available.

<https://docs.prefect.io/2.10.13/concepts/work-pools/#configuring-prefetch>

Thanks for the tip <@U03RN8W7DPU>. It seems our flow runs were transitioning from scheduled &gt; failed &gt; running because prefects k8s service account did not have rights to access `kube-system`  namespace to determine cluster UID. Setting it explicitly via ENV <https://discourse.prefect.io/t/deployment-fails-trying-to-get-namespace/2199/7|as in here> resolved it. The helm charts do not include the necessary rights AFAIK. :thank-you: