• Docker Hub limits unauthenticated users’ image pulls.
• Company struggles to find ways to monetise.
• Automated downloads from Docker Hub mean some users will hit limits, fast.

As of today, April 1 2025, access to containers hosted on Docker Hub will be limited for unauthenticated users or those on the Personal tier (which is free but requires registration).

The change may have significant impact on development pipelines, in particular where CI/CD workflows involve the automated pull of container images from Docker.

While teams may access Docker Hub manually, and some will already be aware of the impending issue, it’s quite common for requests for container images to stem from workflows run by automation software offered by the likes of GitLab and GitHub. Requests made by pipeline automation systems will count towards the rate limit, as such software identifies itself as an unauthenticated user. GitLab’s Dependency Proxy, for example, pulls container images in this manner.

There’s also the issue that many users’ requests stemming from runners (automated development pipeline software) will come from the same IP address, so will hit the limit buffers very quickly. That’s especially true if teams are using hosted runners from providers such as GitLab, but will also be true if CI/CD software is run in-house.

As of April 1, 2025, Docker Hub Personal accounts will be allowed 100 image pulls per hour, while unauthenticated requests are kept down to just 10 per hour. Paying users of Docker get unlimited pull requests, with the usual ‘reasonable use’ caveat.

However, in a statement to the Register, last week, Docker spokesperson David Oro said that the limits would only affect around seven percent of users in total.

The rate limits initially suggested by Docker were set considerably lower than those in force as of today, and the timing of the download cap’s imposition was delayed by a month to give teams extra time to prepare.

What are containers?

Docker has rapidly become one of the big names in containerisation and is credited with developing the technology. Containers allow developers to run applications in what might be thought of as ‘miniature virtual machines.’ Containers usually comprise of a single application or service complete with its dependencies, but no operating system bundled with it. The OS functions are handled by the machine hosting the containers, so container images are smaller and often faster than running a full VM with an application inside it.

The bare bones of applications can be built using multiple containers: one container may hold the main application code and its direct required dependencies. Other containers run alongside, providing, for example, a web server, databases, proxy services that act on traffic to and from the main application, a container that logs network traffic, and so forth.

Developers can therefore build what are often complex code infrastructures by pulling ready-made containers from services like Docker Hub. However, the iterative nature of software development means the regular rebuilding of an application throughout its lifecycle (processes which platforms like GitHub and GitLab can automate). Each rebuild or iteration might request the repeated download of specific containers from repositories like Docker Hub.

It’s easy to see, therefore, that hitting rate limits – if applicable – could disrupt many software development workflows without development teams being particularly aware of the multiple requests for containers that their activities set in motion.

Monetising Docker

Hosting containers and serving them to users costs a great deal in resources, and the Docker company has struggled to find ways to cover its costs, let alone monetise profitably. Privately held Docker is funded by venture capital, whose stakeholders expect returns on their investment.

Over recent years, Docker has floated the idea storage-based charges for its users, but has since said that such plans have been shelved indefinitely. Although the prices Docker charges have risen a couple of times since 2021, industry observers have often expressed puzzlement at exactly how a service like Docker can survive economically.

As is often the case in the open-source software space, there is the misnomer in some organisations that the word ‘free’ in ‘free and open-source’ refers solely to the monetary cost of software to its users. Freely-distributable software (which containers usually…contain(!)) and its distribution methods like Docker Hub are highly valuable. But free-as-in-monetary-value in software is too often considered a right rather than a blessing, and recognition of software’s worth doesn’t necessarily mean people are willing to pay for it.