I run Cilium with default-deny on a few clusters. Every time a new service deploys, something gets blocked, and I end up in the same loop: read Hubble logs, find the dropped flow, write the CiliumNetworkPolicy YAML, apply, wait, repeat.
I wrote CPG to skip that loop. It connects to Hubble Relay, streams denied flows, and generates the policy files. TCP/UDP, ICMP, reserved entities, CIDR — it handles what I've needed so far. If a policy already exists on disk, it merges the new rules in without duplicates.
It's a CLI tool in Go. You can also install it as a kubectl plugin via krew (PR pending). It auto port-forwards to hubble-relay, so there's zero setup beyond having Cilium running.
Built most of this with Claude as copilot, so fair warning on that front.
Would be curious to hear if others have a different approach to the default-deny policy bootstrapping.
I wrote CPG to skip that loop. It connects to Hubble Relay, streams denied flows, and generates the policy files. TCP/UDP, ICMP, reserved entities, CIDR — it handles what I've needed so far. If a policy already exists on disk, it merges the new rules in without duplicates.
It's a CLI tool in Go. You can also install it as a kubectl plugin via krew (PR pending). It auto port-forwards to hubble-relay, so there's zero setup beyond having Cilium running.
Built most of this with Claude as copilot, so fair warning on that front.
Would be curious to hear if others have a different approach to the default-deny policy bootstrapping.
Here is the link : https://github.com/SoulKyu/cpg