Details
-
Task
-
Status: Resolved
-
Critical
-
Resolution: Fixed
-
2.4.0
-
None
Description
Script alerts, such as Ambari Agent / Host Disk Usage needs to have some metrics Externalized.
The Usage screens relies on the alert_disk_space.py script to collect metrics such as Disk utilization (percentage) and directory being checked.
The problem is that the directories being check in this script are:
1. ROOT ("/")
2. /usr/hdp
If ROOT is OK, then "/usr/hdp" is checked subsequently. However if ROOT is NOT OK (>80% utilization) then "/usr/hdp" is ignored and the "CRITICAL" tag applies to ROOT.
This would work on a OOTB standard cluster. However, many users, have Partitions for the hadoop directories. In this instance, "/" is static, no new content will be added. In this instance being at 80 or 90% is OK, this directory will never grow beyond that.
"/usr/hdp" is a separate partition that grows in time.
Based on the above described check, the 80% of the root partition will trigger a "CRITICAL" message, although the "/usr/hdp" partition is OK.
The selection of what directory is being checked should be EXTERNALIZED to a configurable property to avoid these kinds of misleading messages.
CONVERSELY — what constitutes "OK" "WARN" or "CRIT" (the percentage that triggers these) should be left up to the user, not hardcoded in the script.
This is a request to externalize the directories that should be checked, as well as the Percentages into an external configurable property.
Attachments
Attachments
Issue Links
- duplicates
-
AMBARI-16878 Allow for easy, and persistent editing of SCRIPT alert thresholds
- Resolved