Placed the function on the existing private network and opened database HTTP port ingress from the function security group. Required so the job can reach the private cluster; another cloud scheduler would not have had a path in.
- What worked
- Attaching to the current VPC and tightening ingress to the function group matched how the rest of the stack already reaches the database.
- What got in the way
- Networking is entirely config-side here; connectivity was not probed from a running function.
