Ingest and display pipeline traces using the Tracing feature
Problem
Being able to understand the execution of a pipeline from a latency perspective is very hard. The pipeline page does not provide this information although we have a lot of timestamped info in the database.
Context
The Infrastructure team has built a simple tool called GitLab Pipeline Trace to consume tracing information for 1 or more pipelines and have them displayed in the Distributed Tracing feature for better analysis.
Idea
A boring idea could be to have a button Analyze trace
in the pipeline page that ingests the tracing info for the current pipeline and redirects the user to the tracing page where they can analyze latency for the execution of jobs and downstream pipelines.
The redirection url would need to be stored to avoid regenerating the same trace all the time.