RELEASED: Public: April 2, 2025
Resolved issues
- If High Availability (HA) build adoption is slow, permanent agents can be acquired by other builds (BEE-56873)
-
When a High Availability (HA) controller replica exits, any builds it was running are offered for adoption by other replicas. Any agents currently in use by that build remain reserved for it. If the adoption process is slow (for example, if the build record takes a long time to load), a permanent (non-cloud) agent occupied by that build might incorrectly be considered available for other builds. This could result in workspace corruption as unrelated tasks run. If several minutes pass, the original build might abort.
Now, the High Availability (HA) controller records agent usage by specific builds (or build stages), ensuring that other replicas do not allow the agent to be used by any build except the same one once it is adopted.