[YARN-6862] Nodemanager resource usage metrics sometimes are negative

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.8.2
• Fix Version/s: 3.4.0, 3.3.2, 3.2.4
• Component/s: nodemanager
• Labels: pull-request-available
• Target Version/s: 3.3.5
• Hadoop Flags: Reviewed
• Assignee: Benjamin Teke
• Reporter: YunFan Zhou
• Created: 24/Jul/17 10:16
• Updated: 04/Jan/22 22:42
• Resolved: 17/Dec/21 13:54

Description

When we collect real-time metrics of resource usage in NM, we found those values sometimes are invalid.
For example, the following are values when collected at some point:

"milliVcoresUsed":-5808,
"currentPmemUsage":-1,
"currentVmemUsage":-1,
"cpuUsagePercentPerCore":-968.1026
"cpuUsageTotalCoresPercentage":-24.202564,
"pmemLimit":2147483648,
"vmemLimit":4509715456

There are many negative values, there may a bug in NM.
We should fix it, because the real-time metrics of NM is pretty important for us sometimes.

Issue Links

• duplicates: YARN-5021 -1B of 3 GB physical memory used
• links to: GitHub Pull Request #3526
• links to: GitHub Pull Request #3770
• links to: GitHub Pull Request #3771