RunRight

Right-size your CI machines. RunRight runs alongside every CI job, samples CPU/memory/disk/threads, and recommends the cheapest AWS or GCP instance that still fits your workload.

Self-hosted · ELv2 · No SaaS · Container-aware (cgroup v2/v1)

---

Why RunRight

Grafana, Datadog, and Sentry tell you how your application behaves. They do not tell you whether you are paying for the right machine. RunRight solves that: given your actual p95 CPU and memory usage, it recommends the cheapest AWS or GCP instance type that still fits your workload.

Key Features

• No code changes. Runs as a sidecar—not an SDK or agent you instrument. Drop it in any CI pipeline.
• Minimal overhead. Polls /proc every 5s. Footprint: under 5 MB RSS, under 0.1% CPU. Your builds stay fast.
• Actionable insights. Get specific instance recommendations with cost delta and tier labels—no graphs to interpret.
• Works with your stack. Exports via OTLP, Prometheus, or JSON. Integrates with Datadog, Grafana Cloud, or any observability tool.

---

In action

See runright-platform for a live product tour of the self-hosted dashboard.

---

Install

Binary — macOS / Linux

curl -fsSL \
  "https://github.com/gbudjeakp/run-right/releases/latest/download/runright_$(uname -s | tr '[:upper:]' '[:lower:]')_$(uname -m | sed 's/x86_64/amd64/').tar.gz" \
  | tar -xz && sudo mv runright /usr/local/bin/

GitHub Actions — no install needed, use the action directly:

- uses: gbudjeakp/run-right@v1
  with:
    run: make build

Already using Datadog, Grafana Cloud, New Relic, or any OTLP tool? Point RunRight at your existing collector:

OTEL_EXPORTER_OTLP_ENDPOINT=https://your-collector:4317 \
  runright monitor --export otlp --job-id my-build

Using Prometheus?

runright monitor --export prometheus --prometheus-port 9090

---

Add to CI

GitHub Actions

Two usage modes:

Wrapper mode (simplest) — wraps a single command:

- uses: gbudjeakp/run-right@v1
  with:
    run: make build

Standalone mode — monitors across multiple steps:

- uses: gbudjeakp/run-right@v1
  with:
    step: start

- run: make build
- run: make test

- uses: gbudjeakp/run-right@v1
  with:
    step: stop

Send to an OTLP collector (Datadog, Grafana Cloud, New Relic, etc.):

- uses: gbudjeakp/run-right@v1
  with:
    run: make build
    export: otlp
  env:
    OTEL_EXPORTER_OTLP_ENDPOINT: ${{ vars.OTEL_ENDPOINT }}
    OTEL_EXPORTER_OTLP_HEADERS: "Authorization=Bearer ${{ secrets.OTEL_TOKEN }}"

Send to self-hosted RunRight Platform:

- uses: gbudjeakp/run-right@v1
  with:
    run: make build
    export: http
    http-url: ${{ vars.RUNRIGHT_URL }}
  env:
    RUNRIGHT_API_KEY: ${{ secrets.RUNRIGHT_API_KEY }}

Use step outputs (e.g. in a subsequent step or job):

- uses: gbudjeakp/run-right@v1
  id: rr
  with:
    run: make build

- run: echo "Suggested machine: ${{ steps.rr.outputs.suggested-machine }}"

Action inputs

Input	Default	Description
run	—	Command to wrap (wrapper mode)
step	—	start or stop (standalone mode)
export	file	Export backends: file, otlp, http, prometheus (comma-separated)
interval	5s	Metrics sampling interval
duration	0 (unlimited)	Max monitoring duration, e.g. 30m
job-id	${{ github.job }}	Unique identifier for this run
output-dir	.runright/	Directory to write metrics files
http-url	—	RunRight Platform URL for http export
upload-artifact	true	Upload metrics files as a GitHub Artifact
pr-comment	true	Post recommendation as a PR comment
github-token	${{ github.token }}	Token used to post PR comments
provider	(all)	Filter recommendations: aws, gcp, or github
version	latest	RunRight binary version to download

Action outputs

Output	Description
recommendation-json	Full JSON array of machine recommendations
suggested-machine	Top recommended machine ID (e.g. t3.medium)
detected-machine	Machine type detected at runtime
artifact-url	URL of the uploaded metrics artifact

Jenkins

sh '''
  OTEL_EXPORTER_OTLP_ENDPOINT=$OTEL_ENDPOINT \
  $RUNRIGHT_BIN monitor --export otlp \
    --job-id "$JOB_NAME-$BUILD_NUMBER" --interval 3s &
  RR_PID=$!
  make build
  kill $RR_PID
'''

Kubernetes: set pod resource limits and RunRight reads them via cgroup v2:

resources:
  limits:
    cpu: "2"
    memory: "2Gi"
env:
  - name: RUNRIGHT_VCPUS
    valueFrom:
      resourceFieldRef: { resource: limits.cpu }
  - name: RUNRIGHT_MEMORY_GIB
    valueFrom:
      resourceFieldRef: { resource: limits.memory, divisor: 1Gi }

---

Tier	Meaning
right-sized	Current machine fits well
cheaper	A smaller instance meets headroom needs
more-headroom	Current machine is too small

Catalog: 160+ AWS and 60+ GCP instance types, matched at p95 CPU + memory with 20%/30% headroom buffers.

---

CLI

runright monitor   [--duration D] [--interval I] [--export file,http,otlp,prometheus] [--job-id ID] [--http-url URL]
runright recommend [--metrics FILE] [--provider aws|gcp] [--format table|json]
runright catalog list [--provider aws|gcp] [--family NAME] [--arch x86_64|arm64]

---

Dev

make build              # local binary
make build-linux        # linux/amd64 + linux/arm64
make build-all          # all 5 platforms
go test ./internal/...

For the self-hosted dashboard and API, see runright-platform.

---

Elastic License 2.0 (ELv2). See LICENSE