Strict vs DIY Self-Hosted Runners & AWS EC2 Mac
Installing actions-runner takes five minutes. Keeping five self-hosted
Macs healthy for six months is the hard part.
Every team that rolls their own runner scripts hits the same operational bugs:
unallocated APFS container space on EC2 Mac, full DerivedData disks, macOS
sleeping mid-build, silent Xcode version skew across nodes, and dropped runners with no
alert. Strict automates every one of them.
Hardware vs AWS EC2 Mac dedicated hosts
AWS requires a 24-hour minimum dedicated host allocation for EC2 Mac instances—two M1 instances run roughly $1,000/month regardless of utilization.
Spare Mac + Strict vs AWS EC2 Mac
macOS Apple Silicon · 30,000 min/mo
Even if you run EC2 Mac dedicated hosts, Strict still installs cleanly on top of them to solve APFS container sizing, disk bloat, load balancing, and alerting.
Real-world self-hosted failures Strict fixes automatically
| Failure mode | With Strict | DIY Shell Scripts / Raw LaunchDaemons |
|---|---|---|
Full DerivedData / Docker disk |
7-tier auto cleanup down to targetGiB, skipping active
Runner.Worker builds
|
Blind rm -rf cron deletes in-flight build workspace mid-PR
|
| Unallocated EC2 / APFS disk |
strict doctor checks diskutil apfs list free
container space and prints exact fix
|
500 GB EBS volume only exposes 100 GB to macOS until build dies |
| macOS sleep & autologin |
Doctor verifies pmset sleep, autologin plist, and Linux
loginctl linger
|
Runner goes offline overnight or after every OS reboot |
| Xcode / Swift version drift | 12-toolchain inventory on every 30s heartbeat + fleet drift alert | One Mac updates Xcode silently; TestFlight builds fail randomly |
| Hot or overwhelmed node | Auto load-balancer drains slots per repo pool when a healthy peer has headroom | Overloaded node keeps accepting GitHub jobs until OOM/disk full |
| Daemon exits cleanly |
launchctl/systemd user services +
self_heal + before-job.sh barrier hook
|
Runner stays offline until someone SSHes in manually |
| Fleet visibility & alerts | Menu-bar macOS app, web glass console, Cloudflare Email + webhook alerts | Developers notice offline runners only when PRs hang |
Adopt your existing runners in place—no wipe needed
If you already have ~/actions-runner or
~/actions-runner-* configured on a Mac or Linux host, Strict discovers
them automatically, leaves your GitHub registrations intact, wires
ACTIONS_RUNNER_HOOK_JOB_STARTED to before-job.sh, and starts
reporting within 30 seconds.
curl -fsSL https://strict.ai/install.sh | sh
strict doctor
See every cleanup tier and preflight check in detail
Inspect how 7-tier disk reclamation, 9 doctor checks, and risk-aware alerts keep self-hosted runners green.
Frequently asked questions
Why do DIY self-hosted GitHub Actions runners fail silently?
Raw actions-runner setups commonly fail from full DerivedData or Docker disks, macOS sleep timeouts, disabled systemd linger on SSH logout, unexpanded APFS containers on EC2 Mac, DNS flakes, or mismatched Xcode versions across nodes.
Does Strict prevent AWS EC2 Mac's unallocated APFS container trap?
Yes. strict doctor's apfs-unallocated check parses diskutil apfs list automatically and alerts you before jobs crash with false out-of-disk errors—even if you run Strict on EC2 Mac dedicated hosts.
Can Strict replace custom cron cleanup scripts safely?
Yes. strict_maintenance.py runs inside maintenance.lock, skips active Runner.Worker builds, and stops reclaiming as soon as free disk reaches targetGiB across Tiers 1–5.