Created on 07-11-2018 11:52 PM - edited 09-16-2022 06:27 AM
Noticed this in CM 5.13
When changing the value Node Manager config item Dish Health Checker Max Disk Utilization Percent (yarn.nodemanager.disk-health-checker.max-disk-utilization-per-disk-percentage) - in my case increasing it from 90 to 95, CM doesn't flag the already running instances as being out of date/stale and flagged for restart
This is confirmed by looking in the running processes yarn-site.xml file
Also I think the "Dish Health" might mean "Disk"
No further support required here, just pointing it out
Regards
Neil
Created 07-12-2018 08:57 AM
Thank you for bringing this to our attention.
I just tested on CM 5.13.1 and 5.15.0.
I changed the percentage from 90 to 91.
After about 10 seconds, both YARN services showed the configuration was stale. The diff showed the change I made
I wonder if there may have been a problem with the staleness detection in CM in your case.
If this happens again, you might try restarting Cloudera Manager and see if the stale config is detected and then review the CM log to see if there was anything going wrong with "staleness" detection.
Thanks again!
Created 07-12-2018 08:57 AM
Thank you for bringing this to our attention.
I just tested on CM 5.13.1 and 5.15.0.
I changed the percentage from 90 to 91.
After about 10 seconds, both YARN services showed the configuration was stale. The diff showed the change I made
I wonder if there may have been a problem with the staleness detection in CM in your case.
If this happens again, you might try restarting Cloudera Manager and see if the stale config is detected and then review the CM log to see if there was anything going wrong with "staleness" detection.
Thanks again!