Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
-
Mesosphere Sprint 17, Mesosphere Sprint 18, Mesosphere Sprint 19
-
3
Description
The lack of synchronization between ProcessManager destruction and the thread pool threads running the queued processes means that the shared state that is part of the ProcessManager gets destroyed prematurely.
Synchronizing the ProcessManager destructor with draining the work queues and stopping the workers will allow us to not require leaking the shared state to avoid use beyond destruction.
Attachments
Issue Links
- contains
-
MESOS-2879 Random recursive_mutex errors in when running make check
- Resolved
- is related to
-
MESOS-2879 Random recursive_mutex errors in when running make check
- Resolved