Engineer Nick Barrett in its wondered why
Docker Volume Container are now beginning to use even for the decision not to do tasks "podhod shah" tool?
Today Docker Volume Container used for anything and everything, and, according to Barrett, this leads him into confusion.
To illustrate this, an engineer proposes to consider the situation with the launch of the storage of images (Docker Registry) on Docker (v2):
- Run a single instance with Go binary;
- Thus saving disk space capacity;
- The relatively low cost of CPU/memory.
This configuration does not need to Kubernetes Cluster, besides there are not required capabilities to scale, which gives the Docker. So all that can be run directly on the "hardware".
The trouble is that the installation does not find information about the parameters of how this can be done. In fact, the "official" way - use the image Docker. Fortunately, there is a simple script Dockerfile, so that the problem can be solved by using them (the way docker / distribution -> Registry Image -> Dockerfile).
There are other options for work outside Docker. It's about data-stores (temporary storage). Let's say you need to run a cluster or Elasticsearch Galera. Docker-containers offer quick installation that looks pretty tempting.
But no need to hurry. How can I configure these services to work with different media (test / prod clusters)? They can not use variables ENV and do not know anything about the internal service uses the tools of disclosure. These types of systems have their own configuration files, elasticsearch.yml or my.cnf. In this and similar cases Dockerfile format terribly useless, says Barrett.
Engineer sure that the most common solution is easy to install other utilities within the image that would be charged to the configuration before starting the service. And in his opinion, an insult to the very idea of the existence of containers without non-specialized software. Tools, such as pyinfra and Ansible for such work much more convenient (they do not set rubbish to generate the config).
Docker Volume Container
After all, you can use readily available instances Elasticsearch / Galera / etc, which is much more useful at the stage of product development. The ability to quickly run a single instance Elasticsearch-tied to a specific branch applications, save a lot of time. It is definitely the best way to deploy stateless-application ever created. Therefore Barrett advises "just do not bother" with the creation of clusters or integrated foreign software via Docker containers.
