@Marvin Sometimes it takes some time for my workflow to be scheduled onto a pod and for the pod to be ready in the kubernetes cluster. I deployed my prefect infra using helm chart, which values should i change to increase the timeout before prefect marks the workflow as "crashed" when in reality my pod is in the process of being created?

To prevent Prefect from marking your workflow as "crashed" during the pod creation process in Kubernetes, you may consider increasing timeouts that control the scheduling and deployment timeouts. Unfortunately, the initial search didn't return a direct answer to your question. However, you can usually control these types of settings through Helm values when deploying Prefect, specifically looking for configurations related to flow run timeouts, deployment timeouts, or Kubernetes-specific settings that might impact how the Prefect scheduler interacts with your Kubernetes cluster. Since the initial search was not fruitful, I'll now search specifically for Prefect's helm chart configuration which should provide details on the relevant helm values to adjust: 1. Query the Prefect documentation for Helm chart-specific configurations. 2. Get information about Kubernetes Runner in Prefect documentation. Let's proceed with searching again: To ensure that Prefect does not prematurely mark your workflow as "crashed" during the Kubernetes pod creation process, you should look into the “Pod Watch Timeout Seconds” setting in your Helm chart configuration. You can increase this timeout to allow more time for pods to be scheduled and readied on your Kubernetes cluster. Create or modify your

values.yaml

file for the Prefect Helm chart to include an increased value for

podWatchTimeoutSeconds

. For instance:

yaml
worker:
  podWatchTimeoutSeconds: <DESIRED_TIMEOUT_VALUE>

Replace

<DESIRED_TIMEOUT_VALUE>

with an appropriate number of seconds to wait for pod creation. Please refer to the Prefect Kubernetes deployment documentation for more information on customizing the Helm chart values. Note that other relevant configurations might be available; always ensure that your values correspond with the current version of the Helm chart you are using.