@Marvin we have successfully deployed Prefect on k8s with additional worker pods. Works amazing! Unfortunately when I took a look at the dashboard again today it looks like all settings are removed, like the default namespace for the worker pool, and no previous runs are visible, deployments have been removed etc. Looks like a clean install again. We deployed by following https://docs.prefect.io/v3/deploy/infrastructure-examples/kubernetes. How can we fix this?

It seems like you're experiencing an issue where your Prefect deployment on Kubernetes appears to have reset, losing previous configurations and run history. This could be related to a few known issues: 1. Task Runs Missing: There was a known issue from versions 3.0.0rc18 to 3.0.0rc19 where task run completions were not registered correctly, and the dashboard would not display any task runs. If you're using a similar version, consider downgrading or upgrading to a stable release. GitHub Issue #15153. 2. Deployment Configuration: Ensure that your Kubernetes deployment configurations are persistently stored. If your deployment YAML or Helm charts are not configured to persist data, it might appear as a fresh install after a restart or redeployment. 3. Database Connection: Verify that your Prefect server is correctly connected to a persistent database. If the database connection is lost or misconfigured, it might result in the loss of previous runs and configurations. 4. Kubernetes Namespace and Job TTL: Double-check your Kubernetes namespace settings and job TTL configurations. By default, jobs are created in the

default

namespace, and completed jobs are automatically cleaned up after a specified TTL. Ensure these settings are as intended in your deployment configuration. For further guidance, you might want to revisit the Prefect Kubernetes Deployment Guide to ensure all configurations are correctly set. If the issue persists, consider reaching out to the Prefect community or support for more specific troubleshooting based on your setup.