[YARN-10468] TestNodeStatusUpdater does not handle early failure in threads

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.4.0
• Component/s: nodemanager
• Hadoop Flags: Reviewed
• Assignee: Ahmed Hussein
• Reporter: Ahmed Hussein
• Created: 20/Oct/20 20:51
• Updated: 24/Nov/20 19:10
• Resolved: 24/Nov/20 19:10

Description

While investigating HADOOP-17314, I found that the

• TestNodeStatusUpdater#testNMRegistration() will continue running while (heartBeatID <= 3 && waitCount++ != 200) { even though the nm thread could already be dead. the unit should detect that the nm has died and terminates sooner to release resources for other tests.
• TestNodeStatusUpdater#testNMRMConnectionConf(). Same problem as described above.

Issue Links

• links to: GitHub Pull Request #2461