So according to the docs all pipelines run in parallel except those which have “depends_on:”.
In the same time we experience behaviour which is actually contradicts to this logic. Here is the screenshot of our build and graph view:
So as you can see “Deployment” pipeline is skipped although it is not depend on the failed “Custody” pipeline and the they are on the same level of “parallelism”.
For me this sounds like problem which should be very common but I haven’t found anything on the forum so I’m wondering maybe it has something to do with the version.
server - v2.11.1
runner - v1.8.0
Pipelines are scheduled in parallel, however, their ability to execute in parallel depends upon available nodes. If there are not enough nodes to execute all pipelines in parallel, pipelines sit in a pending state in the queue until a node becomes available.
When a build fails, running pipelines are allowed to complete, while pending pipelines are canceled. This means if a pipeline is pending in queue, and the overall build fails, it will not be executed and will have a skip status. So while there may be no relation in the graph, when the build is placed in a failed state, subsequent pipelines are no longer executed (no matter where they are in the graph) unless you explicitly configure them to run on failure.
So based on the above logic, the behavior you are experiencing is not unexpected.