Alright, let's cut through the marketing. Every vendor says they're "great for monorepos" and "scale to any team size." I don't believe it until I see the build minutes on a bill.
We're evaluating a shift. Current state:
* Single monorepo, 100+ devs pushing daily.
* 200+ microservices, shared libraries.
* Main branch builds take ~45 minutes currently (homegrown system).
* Need: fast PR validation (<10 min), efficient main branch builds, sane cost per developer per month.
The shortlist everyone throws out: GitHub Actions, GitLab CI, CircleCI, Buildkite.
I want concrete, billable metrics. Not "feels fast."
* If you're running a similar setup, **show me your actual pipeline configs** for monorepo optimizations. How do you handle selective builds? DAG? I'll start:
```yaml
# Example of a pattern we're testing for selective service builds
# Using changed-files in GitHub Actions to trigger only relevant svc
- name: Get changed services
id: changes
uses: dorny/paths-filter@v2
with:
filters: |
services:
- 'services/my-svc/**'
libs:
- 'libs/shared-lib/**'
```
* **What's your actual cache hit rate?** 90%? 50%? Show the logs.
* **What's your monthly spend on compute for CI/CD?** This is the only number that matters. If you won't share dollars, share build minute consumption.
* How are you handling the queue for 100+ devs? Does your tool become a bottleneck where devs are waiting for an available runner?
I'm inherently skeptical of any "unlimited minutes" plan. They're never truly unlimited for serious workloads. The math always comes out in the bill.
Benchmark numbers I care about:
- PR build time (avg, p95)
- Cost per successful build
- Time from commit to deployed staging (if applicable)
- Maintenance overhead (FTE to manage the CI system)
Anything else is just noise.
show me the bill