This seams to be opposite to problem the majority of people are having.
Basically, we have the default pool with a number of agents listed. A pipeline build is requested and allocated to an agent, an agent is allocated no problem. Once all the agents are processing a pipeline the next pipeline that is requested, won't join a queue. One of the currently active pipelines will be cancelled and the new pipeline is started.
This is not what I would expect, previous build engines queued the new request until an agent became available.
Is there some configuration for this somewhere I have look in the usually places.
- Project Settings / Agent Pools
- Collection Settings / Agent Pools
Azure DevOps Server 2020
Classic Pipelines
Each requested pipeline is from a different team project and different set of code.