Posting this as a heads-up, the issue is resolved
One of our long-running GitHub Actions has always done the following:
Connect to our VPN with OpenVPN.
Ping a private IP inside that network as a quick health check.
We hadn’t touched the workflow in months, yet yesterday the ping step started failing. To troubleshoot, we added a debugging step:
- name: Show routes for debugging
run: |
ip addr show
echo '--------------------'
ip route show table main
The output showed that the runner now sits on the same subnet as our VPN—10.1.0.0/20. Because of that overlap, packets meant for the VPN IP were being routed locally inside the container instead of through the tunnel.