Compute
Containers review
Score: 5/10The missing piece of the platform, trapped inside the platform.
The breakdown explains the score. It is not an average. How to read these.
Last updated
Containers fills an obvious gap in the Workers platform. Sometimes you need a real Linux process for an existing binary, an unsupported runtime, or CPU-heavy work. You can now run one on Cloudflare’s network and address it from a Worker.
That Worker dependency is also the biggest structural drawback. Every container is reached through a Worker and managed by a Durable Object, so you cannot deploy one without buying into the rest of the platform. This is an add-on for Workers architectures, not a standalone container service, and it inherits the operational concerns of Durable Objects. Instance choices are limited, cold starts take seconds, and sustained workloads can cost more than simply renting a VM. That leaves it squarely in the middle.
For heavy, stateful compute, owned hardware still wins comfortably. As burst capacity for a Workers-first architecture, though, Containers is already handy for the odd FFmpeg job, legacy binary, or build step. It is also clearly early in its life.
Status page incidents
Published under Containers, plus incidents whose title names it.
- Last 14 months
- 18
- 4 from the title
- Major or critical
- 2
- Median time open
- 3.0 h
- Since July 2025
- 21
Counts what Cloudflare published, not what broke: time open is how long the status entry stayed open, and nothing before July 2025 is tracked at all. Incidents Cloudflare tagged against nothing but named in the title are counted too. The caveats in full.