What a base image is and why it is refreshed
A container image is built from layers. The lowest layer is the base image with the operating system and the runtime environment, and dependencies and the application code sit above it. Security patches most often appear in the base layers, so if the build always reuses the stored layers, an old base version can stay in the image for months.
That is why the first build of each day on a given runner is made with the base images refreshed and without cache. The goal is twofold: to pick up security patches and to regularly prove that the build path from scratch works. If that path ever fails, it fails on an ordinary working day, not during an urgent fix.
Does the daily build change what runs at the plant
No. The build is produced in the vendor's pipeline and does not change the instances at the client's site. Deployment to a host is a task started manually, so the timing is decided by a person, not by the build schedule.
In addition, deployment cannot be interrupted: a new commit cancels an old, unfinished analysis pipeline, but it never stops a deployment that has already started halfway. For the plant this means an update is not left in an intermediate state because of another change in the repository.
Pinned component versions (major version)
How we pin versions
The versions and base images are pinned: PHP 8.4, MySQL 8.0 and Redis 8.6. Pinning means that changing the major version of a component is a decision and a change in the configuration code, not a side effect of pulling the newest image.
For the IT department this matters during an audit: the list of components and their versions is fixed within a release. Refreshing the base images concerns patches within the pinned version, not a jump to a new major version.
- PHP 8.4
- MySQL 8.0
- Redis 8.6
Why layer caching shortens the download
Apart from the daily build without cache, the opposite rule applies: dependency installation is tied to the lock files, and the source code is copied as the last layer. Hosts then download only the changed layers instead of the whole image of about 357 MB.
In an on-premise installation this translates into less traffic through the handover point and a shorter maintenance window. A change to the code alone does not force a new download of the dependency layers.
What monitoring should show after an update
After a deployment it is worth checking three things: the image version on the host, the status of the containers and the result of a simple login test. Recording these three values in the change log gives audit evidence and makes a rollback to the previous version easier.
Before updating production, the same change should go through the twin pre-production instance. The refreshed base layers are then verified on a copy of the environment before they touch the system on which the work permits depend.
Video
Frequently asked questions
Does refreshing the base images restart the system at the plant?
No. The refresh takes place in the build pipeline. A change on the host happens only after the deployment is started manually.
Can an update be interrupted halfway?
A deployment to the host that has started cannot be interrupted. A new commit cancels only an old, unfinished analysis pipeline.
Will a new major database version appear on its own?
No. The versions are pinned, so changing the major version is a deliberate change of the configuration.
Why build without cache once a day?
To pick up security patches from the base layers and to regularly prove that building from scratch works.
Let's talk about the installation in your server room
We will present the requirements and the split of responsibilities for your plant based on the installation documentation. Book a conversation with the InnerWeb team.
Let's talk