[YARN-10448] SLS should set default user to handle SYNTH format

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.2.1, 3.4.0
• Fix Version/s: 3.4.0
• Component/s: scheduler-load-simulator
• Hadoop Flags: Reviewed
• Assignee: Qi Zhu
• Reporter: Qi Zhu
• Created: 24/Sep/20 09:23
• Updated: 13/Oct/20 15:55
• Resolved: 13/Oct/20 15:55

Description

When using the synthetic generator json file example from the doc ( https://hadoop.apache.org/docs/current/hadoop-sls/SchedulerLoadSimulator.html#SYNTH_JSON_input_file_format ), it throws the following exception:

java.lang.IllegalArgumentException: Null user
at org.apache.hadoop.security.UserGroupInformation.createRemoteUser(UserGroupInformation.java:1269)
at org.apache.hadoop.security.UserGroupInformation.createRemoteUser(UserGroupInformation.java:1256)
at org.apache.hadoop.yarn.sls.appmaster.AMSimulator.submitReservationWhenSpecified(AMSimulator.java:191)
at org.apache.hadoop.yarn.sls.appmaster.AMSimulator.firstStep(AMSimulator.java:161)
at org.apache.hadoop.yarn.sls.scheduler.TaskRunner$Task.run(TaskRunner.java:88)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)

So the solution is either:
1) to make user_name a mandatory field, or
2) to set default user in SLS code if the json file does not define it.

IMO, solution 2 might be better, because in most cases (if not all) user_name has no impact on scheduler performance, thus it is reasonable to make it an optional field, which is also consistent with the job.user field in SLS JSON file.