Skip to content

Installation

With the operator

One manifest installs the RupturaInstance resource definition, the operator and its permissions in ruptura-system.

oc apply -f https://raw.githubusercontent.com/benfradjselim/ruptura/main/install/operator.yaml
oc -n ruptura-system rollout status deploy/ruptura-operator

Create an instance in any namespace:

apiVersion: ruptura.io/v1alpha1
kind: RupturaInstance
metadata:
  name: shop
  namespace: ruptura-demo
spec:
  level:
    scope: workload
    selector:
      matchLabels: {team: shop}

oc get rupturainstance -n ruptura-demo

The instance reaches Running in about a minute. The operator creates, in the namespace of the instance: the engine Deployment, a Service, a PersistentVolumeClaim, an API key Secret and a ServiceAccount. On OpenShift it also creates a Route, a ConfigMap filled with the service CA, and bindings to cluster-monitoring-view so the engine reads cluster metrics through Thanos.

An invalid instance deploys nothing: its status reports phase Failed with every problem found.

Deleting the instance removes everything the operator created, cluster bindings included.

While the engine is not ready, status.message gives the cause Kubernetes reports: a failing container with its last log line, a pod the scheduler cannot place, an unbound volume claim, or a pod refused at creation. oc get rupturainstance shows it in the Message column.

One operator manages an instance at a time. It holds a Lease named <instance>-operator in the namespace of the instance, and status.operator names its namespace. A second operator on the same cluster leaves the instance alone, and takes it over only after the holder stops renewing for two minutes. Operators older than 1.4.0 do not use the Lease: remove them before installing 1.4.0.

Upgrading to 1.4.2

  • The engine pod meets the Pod Security restricted profile: seccomp RuntimeDefault, all capabilities dropped, read-only root filesystem. It restarts once after the upgrade.
  • status.conditions reports Available, Progressing and Degraded, read by the OpenShift console, Argo CD and kstatus.
  • The operator no longer lists or watches Secrets across the cluster.
  • The web console image listens on 8080 with any user ID and proxies /api/ to RUPTURA_BACKEND_URL.

Upgrading to 1.4.0

  • OTLP over HTTP moved from port 4317, paths /otlp/v1/*, to the standard port 4318, paths /v1/metrics, /v1/logs, /v1/traces. Port 4317 now serves OTLP gRPC. Point otlphttp exporters at :4318, and otlp exporters at :4317.
  • Prometheus remote write is received at :4318/api/v1/write.
  • The engine pod restarts once after the upgrade.

On Kubernetes the operator installs the same way. Every failure mode except crash_loop reads Prometheus, so declare its query endpoint:

spec:
  input:
    prometheus:
      url: http://prometheus.monitoring.svc:9090

Kubernetes, with Helm

helm install ruptura oci://ghcr.io/benfradjselim/charts/ruptura \
  --namespace ruptura-system --create-namespace \
  --set apiKey=$(openssl rand -hex 32) \
  --set ui.enabled=true \
  --set thanos.url=http://prometheus.monitoring.svc:9090

thanos.url is your Prometheus or Thanos query endpoint. With --set openshift.enabled=true the chart uses the OpenShift Thanos querier instead.

Locally

git clone https://github.com/benfradjselim/ruptura && cd ruptura
docker compose up --build

Open http://localhost:8081 and sign in with password ruptura-local-dev.

Images

Component Image
Engine docker.io/selimbf/ruptura
Operator ghcr.io/benfradjselim/ruptura-operator
UI ghcr.io/benfradjselim/ruptura-ui
Helm chart oci://ghcr.io/benfradjselim/charts/ruptura