reported by QE, I think this is probably a web ui not updating issue, but not positive.
1. The property yarn.nodemanager.health-checker.interval-ms on the RM and the DN are set to 135000 ms, or 2 minutes and
$ grep -A 2 "health-checker.interval-ms" yarn-site.xml
2. According to the web ui (e.g http://rm:8088/cluster/nodes), the 'last health-update'
column does not get updated, except when the RM is restated.