Skip to content

getting-started.md

Latest commit

 

History

History
157 lines (112 loc) · 3.94 KB

getting-started.md

File metadata and controls

157 lines (112 loc) · 3.94 KB

Getting started

Requirements

  • Kubernetes 1.28+
  • Rook-Ceph operator deployed
  • Prometheus (scrapes metrics)
  • cert-manager (ops-log webhook needs TLS)
  • kubectl with cluster access
  • RabbitMQ (only for ops-log audit trail)

Container images

Published to GitHub Container Registry on every tagged release:

Image What it does
ghcr.io/cobaltcore-dev/prysm:<tag> Main binary -- runs all producers
ghcr.io/cobaltcore-dev/prysm-webhook:<tag> Ops-log mutating webhook

Tags follow semver (0.0.36), major.minor (0.0), and git SHA.

Latest version: releases.

Producers

Prysm has three producers. Each runs as a separate Kubernetes workload:

Producer Command K8s pattern External deps
RadosGW Usage remote-producer radosgw-usage Deployment RadosGW Admin API
Disk Health local-producer disk-health-metrics DaemonSet smartctl, nvme-cli (bundled)
Ops Log local-producer ops-log Sidecar (via webhook) RGW ops-log file; RabbitMQ (optional)

Quick start

1. RadosGW Usage producer

Pulls bucket/user usage metrics from the RadosGW Admin API.

kubectl apply -f radosgw-usage-deployment.yaml

Full walkthrough: RadosGW Usage.

2. Disk Health producer

Reads SMART/NVMe attributes on every node running Ceph OSDs.

kubectl apply -f diskhealthmetrics-serviceaccount.yaml
kubectl apply -f diskhealthmetrics-daemon-set.yaml

Full walkthrough: Disk Health.

3. Ops Log producer

A mutating webhook injects a sidecar into RGW pods. The sidecar parses RGW operation logs and exposes Prometheus metrics. It can also publish CADF audit events to RabbitMQ.

kubectl apply -f webhook-namespace.yaml
kubectl apply -f webhook-cert.yaml
kubectl apply -f webhook-deployment.yaml
kubectl apply -f webhook-service.yaml
kubectl apply -f webhook-config.yaml

Full walkthrough: Ops Log.

Configuration

All producers accept configuration three ways. In Kubernetes, environment variables are the simplest.

Environment variables (preferred in K8s)

Set via env, envFrom with a ConfigMap or Secret in your manifests.

CLI flags

prysm remote-producer radosgw-usage --admin-url "http://..." --access-key "..." --secret-key "..."

Config file (local multi-producer mode)

prysm local-producer use-config --config=/path/to/config.yaml

See examples/config/config.yaml for the format.

Logging

All producers log structured JSON via zerolog.

Set verbosity with the -v flag:

-v debug    # verbose
-v info     # standard
-v warn     # default
-v error    # errors only

Prometheus integration

Every producer exposes metrics on an HTTP port (default 8080; ops-log sidecar uses 9090).

ServiceMonitor example

apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: prysm-radosgw-usage
  namespace: rook-ceph
  labels:
    prometheus: kube-prometheus
spec:
  selector:
    matchLabels:
      app: radosgw-usage-exporter
  endpoints:
    - port: metrics
      interval: 60s
      path: /metrics

PodMonitor example (sidecar metrics)

apiVersion: monitoring.coreos.com/v1
kind: PodMonitor
metadata:
  name: prysm-ops-log
  namespace: rook-ceph
  labels:
    prometheus: kube-prometheus
spec:
  selector:
    matchLabels:
      app: rook-ceph-rgw
  podMetricsEndpoints:
    - port: prysm-metrics
      interval: 60s
      path: /metrics

Match labels, namespace, and interval to your Prometheus operator setup.

Next steps