docker - Should I loop a container or loop inside a container? - Stack Overflow

admin管理员组

文章数量:1393556

I want to call genetic variants with DeepVariant on an HPC for about 1000 cereal lines. I successfully ran DV for one line with the docker image they provide using Apptainer/Singularity, but for the full set I want to do that automatically.

Now my question, if I build a script to call every line, does it make a difference to start the container for each iteration or should I make my script so that the container is started once and DV called inside for each line?

I want to call genetic variants with DeepVariant on an HPC for about 1000 cereal lines. I successfully ran DV for one line with the docker image they provide using Apptainer/Singularity, but for the full set I want to do that automatically.

Now my question, if I build a script to call every line, does it make a difference to start the container for each iteration or should I make my script so that the container is started once and DV called inside for each line?

• docker
• hpc
• singularity-container
• apptainer

Share Improve this question Follow asked Mar 12 at 8:58 skranzskranz 6511 silver badge1010 bronze badges

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 0

If at all possible, your program should avoid directly interacting with the container system. If you have the option to start the service once and make a series of requests to it, that is probably better than repeatedly starting and stopping the container.

From a development point of view, things are generally a little easier if you don't have a hard dependency on Docker. This will mean it's easier to run your program in environments without Docker (brand-new developer systems, your CI environment) and you can more easily switch to non-Docker setups (Kubernetes, the service running remotely on a big remote host without a container).

There are two big problems with trying to use the Docker API to directly manage a container (or running docker CLI commands):

1. It's all but trivial to docker run a container that takes over the entire host system: it is a huge security risk.
2. This setup would be very specifically tied to Docker proper, and you'll have to write different orchestration code if you want to run something similar in a clustered environment like Kubernetes, or work nicely in a Compose-based setup.

The best case here is to launch your service dependency just once, outside your application code, and use something like an HTTP client library to send calls to it. This would be similar to how you use a containerized database: you (or a Compose file) creates a database container, and without doing anything Docker-specific, your application uses an ordinary database client to talk to it. Avoid anything that directly uses the Docker socket.

本文标签： dockerShould I loop a container or loop inside a containerStack Overflow