Details
-
Task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
Right now the scheduler uses the (synchronous, sequential) ShutdownRegistry which can lead to deadlocks at shutdown. Migrate to (asynchronous, parallel) ServiceManager [1,2] and add an endpoint to inspect the state of individual services running within the scheduler.
[1] https://code.google.com/p/guava-libraries/wiki/ServiceExplained
[2] http://docs.guava-libraries.googlecode.com/git/javadoc/com/google/common/util/concurrent/ServiceManager.html
Attachments
Issue Links
- is related to
-
AURORA-327 Aurora shutdown blocks when attempting to abdicate leadership and zk ensemble is down
- Resolved
-
AURORA-325 Scheduler shutdown commands should execute asynchronously
- Closed