From pipeline tags to database

Got rid of the git tags usage in pipelines. The idea of ​​binding pipelines to tags was interesting, but it turned out to be suboptimal, because in this way, even taking into account periodic automatic cleaning, several hundred THOUSAND tags could accumulate in the client repository, which greatly slowed down GitLab. Instead of tags, we switched to using the already existing postgresql database in the accounting. The script, instead of using a text database with access through a lock file, has also been updated to use the postgresql database, and now the jobs run much faster. The above-mentioned changes made it possible to make the jobs more stable and without large delays in the launch, and their search became easier.

Share this post