Details
-
Bug
-
Status: Resolved
-
Blocker
-
Resolution: Fixed
-
None
-
Mesosphere Sprint 2018-22, Mesosphere Sprint 2018-23, Mesosphere Sprint 2018-24
-
2
Description
Since a resource provider and its operation status update manager run in different actors, for a completed operation, its `UPDATE_OPERATION_STATUS` call may race with an `UPDATE_STATE`. When the `UPDATE_STATE` arrives to the agent earlier, the total resources will be updated, but the terminal status of the completed operation will be ignored since it is known by both the agent and the resource provider. As a result, when the `UPDATE_OPERATION_STATUS` arrives later, the agent will try to apply the operation, but this is incorrect since the total resources has already been updated.