Analyzing Azure Pipelines Agent Pool Usage via the Historical Graph and TaskAgentPoolSizeSnapshot

Why the Answer Is Correct

Agent pool usage is a distinct measurement from pipeline execution, and Azure DevOps supports two complementary ways to examine it. The first is the historical graph for the agent pools, which visualizes how the pool has been consumed over time without any query writing. The second is querying the TaskAgentPoolSizeSnapshot entity, which is the Analytics entity that records agent pool size and usage, so it supports deeper analysis and filtering beyond the built-in graph. Together they cover both the out-of-the-box reporting path and the data path. The vote was unanimous at 100 for A and C. p2006 linked the pool consumption report documentation, and UrbanRellik plus 8fc2e85 both linked the Analytics entity reference page identifying the TaskAgentPoolSizeSnapshot entity set, which is exactly what option C names.

Why the Other Options Are Wrong

Querying the PipelineRun endpoint (D) or reviewing the Pipeline duration report (B) both measure pipeline activity rather than pool consumption. The Pipeline duration report answers how long pipelines and their stages took to run, and the PipelineRun entity records the individual pipeline runs with their outcomes and timing. Neither captures how many agents were available or how the pool was consumed, which is the specific subject of the question. These are useful reports for a different question, namely where pipeline time is being spent, but they cannot show pool usage.

Community Comment Notes

The community was unanimous at 100 for A and C, and the three substance comments converge on the same two sources. UrbanRellik and 8fc2e85 both cited the Analytics entity reference page and pointed out that TaskAgentPoolSizeSnapshot is the entity set backing the pool size and usage data, which is the documentary basis for option C. p2006 cited the agent pool consumption report page, which is what backs the historical graph in option A. The pairing is therefore not a guess but the two documented access paths to the same measurement.