Details
-
Bug
-
Status: Resolved
-
Resolution: Duplicate
-
None
-
None
-
None
-
None
Description
There are three new rest services, one to shutdown just the RM but leave the executor(s) and tasks running, one to shutdown the entire framework by stopping the driver and exiting the RM, and one that shuts down the framework in a more graceful manner than the previous shutdown by stopping all the tasks, the executors, aborting the driver, and stopping the RM.
The shutdown with stopping the driver forces the user to either have another RM in fail over mode or start a scheduler such that it re-registers. The shutdown by aborting the driver allows the RM to be restarted as if nothing had happened. This is great for testing.
A Help menu was added to explain the menu and tab options. And the "About Myriad" was moved here which brings it more in line with industry standards.
Each Shutdown command will bring up a dialog to ask for confirmation and also again explains what is about to happen.
This address [issue 55] (https://github.com/mesos/myriad/issues/55)
The only thing I was not sure about was if part of the idea was to take down all of hadoop or just what was under Myriad control. I implemented the latter. And what I mean by the hadoop would be the hadoop namenode and datanode...would this be considered part of the Yarn cluster? Or is it what I originally assumed the RM, NM, and NMs tasks?
Attachments
Issue Links
- duplicates
-
MYRIAD-55 Add a API to shutdown Myriad
- Resolved