A Kubernetes distribution for production workloads
Point it at bare machines and one command builds the whole cluster with its storage and databases and secrets and policy and observability.
Counted from the repository
- 70alert rules
- 21network policies on a cluster wide default deny
- 13Kyverno policies
- 9validating admission policies
- 24/24images pinned by digest
- 14namespaces
- 5558lines of Go in the failover controller
- 5331lines of tests for it
- 13TLA+ specs checked in 67 model configurations
Each number links to the files it was counted from. They are recounted on every deploy.
What one command builds
Each layer sits on the one below and assumes the one above can be broken into.
- RuntimeFalco watches every node for shells and odd file reads. Reloader picks up config changes.
- ObservabilityMetrics and logs and traces with alerting from the first boot.
- SecretsOpenBao hands each pod its own secrets as files in memory. The app's database password rotates every day.
- DataThree PostgreSQL clusters. A primary with synchronous replication plus a live replica and one that runs five minutes behind.
- StorageEvery volume encrypted with LUKS. Object storage for backups encrypted at rest.
- AdmissionOnly allowlisted images pinned by digest can run. Pod Security and Kyverno check everything else.
- NetworkDefault deny everywhere. Every plain HTTP call is checked at layer 7 and traffic between nodes is encrypted.
- KubernetesRKE2 with the CIS profile and audit logging and secrets encrypted in etcd.
Run it
Three Debian 12 machines with 4 vCPU and 12 GB of RAM each. SSH key access and Ansible on the machine you deploy from.
Deploy your first tenantgit clone https://github.com/sassendes/bedrock.git
cd bedrock
sudo ./tenant-workload/bootstrap/dependencies/ansible_deps.sh
cp tenant-workload/bootstrap/hosts.example.ini tenant-workload/bootstrap/hosts.ini
make deploy