I ran 29 real docker-compose projects on Apple's container runtime. Here's what broke

Short summary

The author tested 29 real-world docker-compose projects from Docker's awesome-compose repo against Apple's new container runtime (v1.0 on macOS 26). 18 of 29 stacks ran successfully — 14 with zero changes and 4 after a one-line compose tweak — while failures were mostly due to upstream sample rot, Docker socket dependencies, or Linux kernel access requirements. Key runtime differences include no volume seeding, cgroup-sniffing JVM crashes, and per-container micro-VM isolation, though Apple's runtime wins on idle memory (~58 MB vs ~7.8 GB VM) at the cost of slower container starts.

• •18 of 29 real docker-compose projects ran on Apple's container runtime, 14 with zero changes
• •Main failures: Postgres volume mount issue (one-line fix), Docker socket dependencies, upstream sample rot, and cgroup-sniffing JVM crashes
• •Apple container uses ~58 MB idle vs Docker Desktop's ~7.8 GB VM, but container start is slower (~0.81s vs ~0.19s) due to per-container micro-VM boot

Generated with AI, which can make mistakes.