Self-Hosted GitHub Actions Runners for Monorepos
Short answer: Monorepo matrix jobs hammer individual build machines with uneven RAM and disk bursts. Strict runs supply-side load balancing (computeFleetDirective), proportional shedding, and Tailnet build caching without proxying GitHub webhooks.
Balance multi-slot monorepo bursts across Macs and Linux boxes with hysteresis, spillover, and shared Tailnet caching.
Drop-in workflow YAML & zero-sudo enrollment
Enroll your host with strict add-runner and route jobs directly via GitHub's native runs-on labels:
jobs:
affected:
strategy:
matrix:
package: [web, api, cli]
runs-on: [self-hosted, Linux, X64]
curl -fsSL https://strict.ai/install.sh | sh
strict add-runner owner/monorepo --slots 4 --labels self-hosted,Linux,X64
How Strict keeps Monorepos runners healthy
| Subsystem | How it applies to Monorepos |
|---|---|
| Disk Cleanup | Tiers 1–3 automatically reclaim runner logs, package manager stores, and git-ignored build artifacts between heavy matrix bursts. |
| Preflight Doctor | runner-hooks ensures every runner slot honors before-job.sh drain barriers so drained slots never steal work from healthy peers. |
| Tracked Toolchains | node, bun, rust, go, python, swift, docker, gh, runner |
| Cache & Balancing | With multi-slot hosts, Strict applies hysteresis thresholds (RAM >= 85% enter / <= 75% exit; disk; CPU; load-skew >= 1.5x peer) and proportional half-shedding while preserving repo-pool coverage. |
When Strict fits vs. when hosted runners win
Use Strict for Monorepos when
- Zero central GitHub webhook proxy—GitHub assigns jobs natively while Strict shapes healthy slots
- Shared Tailnet Build Cache accelerates Turborepo, Nx, Bazel, and Cargo remote caching at LAN speeds
- Formally TLA+-modeled drain/recovery protocol prevents deadlocks during offline spikes
Use cloud-hosted runners when
- 50+ simultaneous matrix shards needed instantaneously for 2 minutes a day
- Public monorepos accepting external fork pull requests
Frequently asked questions
How does Strict load-balance without acting as a GitHub webhook proxy?
Strict runs supply-side load balancing: computeFleetDirective calculates per-host slot drain directives from live telemetry, and before-job.sh pauses/drains overloaded slots so GitHub naturally routes queued jobs to healthy machines.
Can Strict guarantee a repository pool never loses its last healthy runner?
Yes. computeFleetDirective enforces repo-pool coverage guarantees and pauses shedding when no healthier matching host exists in the same pool.
Does Strict prevent oscillation when multiple machines are busy?
Yes. Separate enter/exit hysteresis thresholds, 1.5x load-skew margins, and deterministic tie-breaking keep slots stable instead of flapping.
Try Strict on one spare Mac in under two minutes
Signed Apple Silicon app or single-line CLI · Zero accounts required