Motivation
because of reasons my gitlab-ci job times out during / towards the end of a docker compose up, which takes endless time (network, image size ...)
Problem
All the timeout settings don't have an effect; neither gitlab job timeout, nor docker compose timeout setting.
Solution
configuring the docker daemon helped.
In a dind scenario it is possible to inject configuration to the entrypoint, especially dramatically increasing the retries helped:
services:
- name: docker:20.10.16-dind
entrypoint: [
"dockerd-entrypoint.sh",
"--max-concurrent-uploads=2",
"--max-concurrent-downloads=2",
"--max-download-attempts=20"
]