Deployment Guide

Rastir ships with production-ready deployment configs for local development, three major clouds, and Kubernetes.

Stack Components

Every deployment target deploys the same four components:

Component Role
Rastir Server Receives telemetry, exposes /metrics, serves OTLP
OTel Collector Forwards traces to the trace backend
Prometheus Scrapes /metrics from Rastir, evaluates SRE rules
Grafana Pre-provisioned dashboards + datasources

The only thing that varies per target is the trace backend:

Target Trace Backend
Local (Docker Compose) Tempo (deployed as 5th container)
AWS X-Ray (managed)
Azure Application Insights (managed)
GCP Cloud Trace (managed)
Kubernetes Configurable: Tempo (default), X-Ray, App Insights, or Cloud Trace

Docker Compose (Local)

The fastest way to get the full stack running locally.

Prerequisites

  • Docker and Docker Compose

Deploy 

cd deploy/docker
./deploy.sh          # brings up all 5 services

What’s Running

Service URL
Rastir Server http://localhost:8080
Grafana http://localhost:3000 (admin/admin)
Prometheus http://localhost:9090
Tempo http://localhost:3200
OTLP gRPC localhost:4317
OTLP HTTP localhost:4318

Commands 

./deploy.sh          # start
./deploy.sh down     # stop
./deploy.sh logs     # tail logs
./deploy.sh restart  # restart all

AWS (ECS Fargate)

Deploys to ECS Fargate via Terraform. Traces go to AWS X-Ray via ADOT (AWS Distro for OpenTelemetry).

Prerequisites

  • Terraform >= 1.5
  • AWS CLI configured with sufficient permissions
  • A VPC with subnets

Deploy 

cd deploy/terraform/aws
cp terraform.tfvars.example terraform.tfvars
# Edit terraform.tfvars with your VPC, subnets, etc.
./deploy.sh          # terraform init + apply

Architecture

  • Rastir Server + ADOT — single Fargate task, ADOT as sidecar
  • Prometheus — Fargate task with EFS for persistent TSDB storage
  • Grafana — Fargate task
  • Service Connect — internal service discovery (services find each other by name)

Key Variables

Variable Description Default
vpc_id VPC to deploy into (required)
subnet_ids Subnet IDs for Fargate (required)
rastir_image Rastir server container image ghcr.io/skamalj/rastir-server:latest
rastir_cpu / rastir_memory Server resources 512 / 1024
grafana_admin_password Grafana admin password admin

Commands 

./deploy.sh          # apply
./deploy.sh plan     # plan only
./deploy.sh destroy  # tear down

Azure (Container Instances)

Deploys to Azure Container Instances via Terraform. Traces go to Application Insights.

Prerequisites

  • Terraform >= 1.5
  • Azure CLI (az login)
  • An Application Insights connection string

Deploy 

cd deploy/terraform/azure
cp terraform.tfvars.example terraform.tfvars
# Edit: set appinsights_connection_string, location, etc.
./deploy.sh

Architecture

  • Rastir Server + OTel Collector — ACI container group, collector as sidecar
  • Prometheus — ACI container group with Azure File Share for storage
  • Grafana — ACI container group
  • VNet — all groups communicate over a private VNet

Key Variables

Variable Description Default
location Azure region eastus
appinsights_connection_string App Insights connection string (required)
grafana_admin_password Grafana admin password admin

GCP (Cloud Run)

Deploys to Cloud Run via Terraform. Traces go to Cloud Trace.

Prerequisites

  • Terraform >= 1.5
  • gcloud CLI authenticated
  • A GCP project with Cloud Run and Cloud Trace APIs enabled

Deploy 

cd deploy/terraform/gcp
cp terraform.tfvars.example terraform.tfvars
# Edit: set project_id, region, etc.
./deploy.sh

Architecture

  • Rastir Server + OTel Collector — Cloud Run multi-container service
  • Prometheus — GCE VM with persistent SSD (Cloud Run is stateless)
  • Grafana — Cloud Run service
  • VPC Connector — allows Cloud Run to reach the Prometheus VM

Key Variables

Variable Description Default
project_id GCP project ID (required)
region GCP region us-central1
grafana_admin_password Grafana admin password admin

Kubernetes (Helm)

A Helm chart that works on any Kubernetes cluster (EKS, AKS, GKE, on-prem). By default traces go to Tempo; set traceBackend to switch.

Prerequisites

  • kubectl connected to a cluster
  • Helm 3

Deploy 

cd deploy/k8s
./deploy.sh                                  # default: Tempo
./deploy.sh --set traceBackend=xray          # AWS X-Ray
./deploy.sh --set traceBackend=appinsights   # Azure
./deploy.sh --set traceBackend=cloudtrace    # GCP

Access 

kubectl port-forward svc/grafana 3000:3000 -n rastir
kubectl port-forward svc/rastir-server 8080:8080 -n rastir
kubectl port-forward svc/prometheus 9090:9090 -n rastir

Key Values

Value Description Default
traceBackend tempo, xray, appinsights, cloudtrace tempo
rastirServer.image.tag Rastir image tag latest
prometheus.storage.size Prometheus PVC size 50Gi
grafana.adminPassword Grafana admin password admin

Uninstall 

./deploy.sh uninstall

Directory Structure 

deploy/
  docker/                         # Docker Compose (local)
    docker-compose.yml
    deploy.sh
    otel-collector-config.yaml
    prometheus.yml
    tempo.yaml
    provisioning/                 # Grafana datasources + dashboards
    dashboards -> ../../grafana/dashboards  (symlink)
    rules/                        # SRE rules + alerts (symlinks)

  terraform/
    modules/                      # Shared Terraform modules
      rastir-server/
      otel-collector/
      prometheus/
      grafana/
    aws/                          # ECS Fargate + ADOT → X-Ray
    azure/                        # ACI + OTel → App Insights
    gcp/                          # Cloud Run + OTel → Cloud Trace

  k8s/                            # Helm chart
    Chart.yaml
    values.yaml
    deploy.sh
    templates/

Rastir — LLM & Agent Observability Library