We have a GKE application container running Django that connects to a Postgres database on Cloud SQL via Private IP. The application is configured to run 10 django processes per pod. Whenever a new pod is spun up (either when deploying new code or scaling to meet load) each of the 10 django processes in that pod encounters exactly one error (varies, but always database-related) when connecting to the database with all subsequent database requests being fine. We suspect that the problem is on the Django side and the requests that error do not even make it to Cloud SQL.

How do I view the network requests between the application and Cloud SQL?