|
|
|
|
|
by shykes
4589 days ago
|
|
You can assign permanent names to containers. We're designing the naming syatem to make it completely OK to keep persistent containers. They're just named directories, docker will never remove them on its own. For example a common pattern is to create a placeholder database container with reference dara as a volume, but no actual process running and no network ports. Then successive versions of the database container are started on the side, with shared access to the placeholder's volume. Later you might run a backup container on the same volume. In other words you can point to a particular dataset as a container of its own, separate of the various applications which might access it. All of these interactions are visible to docker so it can authenticate, restrict, log, or hook them in all the standard ways. In fact internally images and containers are stored side-by-side. In future versions we are going to accentuate that similarity. |
|