@Marvin prefect-server's UI becomes "frozen" with 16 concurrent flow-runs distributed across multiple nodes in my kubernetes cluster. i am running prefect-server in my kubernetes cluster with the bundled postgres db via the prefect helm chart. i have tried experimenting with different PREFECT_SQLALCHEMY_POOL_SIZE and PREFECT_SQLALCHEMY_MAX_OVERFLOW values as well as setting the concurrency limit to 8 and 16 on my work pool but the UI still becomes unusable. I am using prefect version 2.18.3. Is there a setting that I am missing?

Your current setup seems to be having difficulties handling the load you're placing on it. While I found a few potential pointers, a comprehensive solution might involve several considerations: 1. Database Settings: Adjusting

PREFECT_SQLALCHEMY_POOL_SIZE

and

PREFECT_SQLALCHEMY_MAX_OVERFLOW

is a step in the right direction. Make sure these values align with your workload requirements and PostgreSQL configuration. 2. Concurrency Limits: It looks like you have already set concurrency limits on your work pool. Ensure the limits you set are applied correctly and are suitable for your system's capabilities. 3. Resource Allocation: Ensure that your Kubernetes cluster has sufficient resources (CPU, memory) allocated to Prefect Server and the PostgreSQL database instance. This might involve ensuring proper resource requests and limits are set within your Kubernetes manifests or Helm charts. 4. Scaling and Load Distribution: Distribute your load across multiple nodes effectively. Properly load balance your workflows to ensure no single node is overwhelmed. 5. Prefect's Compatibility with Dependencies: Check any issues related to dependencies, as highlighted by the issues on GitHub (e.g., certain Python dependencies might lead to deadlocks). 6. Advanced Tuning: Dive deeper into Prefect's advanced configurations, possibly adjusting parameters related to database connections, API responsiveness, and flow execution environments. Given your setup and requirements, it may be beneficial to conduct a deeper investigation into resource utilization on your Kubernetes cluster while running the flows. References: - Guide on using Prefect with Kubernetes: Prefect Kubernetes Guide - Global Concurrency Limits: Prefect Concurrency Limits - Issues related to similar problems can be found here (search results): ModuleNotFoundError on concurrent flows and Deadlock when running deployments It might be helpful to consider reaching out on Prefect's community forums or Slack for more specific, tailored advice.

is there a way to turn off the graph in the UI for a flow run?