[HDFS-15215] The Timestamp for longest write/read lock held log is wrong

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.3.0
• Hadoop Flags: Reviewed
• Assignee: Toshihiro Suzuki
• Reporter: Toshihiro Suzuki
• Created: 10/Mar/20 07:39
• Updated: 24/Mar/20 22:06
• Resolved: 24/Mar/20 21:50

Description

I found the Timestamp for longest write/read lock held log is wrong in trunk:

2020-03-10 16:01:26,585 [main] INFO  namenode.FSNamesystem (FSNamesystemLock.java:writeUnlock(281)) - 	Number of suppressed write-lock reports: 0
	Longest write-lock held at 1970-01-03 07:07:40,841+0900 for 3ms via java.lang.Thread.getStackTrace(Thread.java:1559)
...

Looking at the code, it looks like the timestamp comes from System.nanoTime() that returns the current value of the running Java Virtual Machine's high-resolution time source and this method can only be used to measure elapsed time:
https://docs.oracle.com/javase/8/docs/api/java/lang/System.html#nanoTime--

We need to make the timestamp from System.currentTimeMillis().

Issue Links

• is related to: HDFS-14775 Add Timestamp for longest FSN write/read lock held log
• links to: GitHub Pull Request #1889