This is my first time posting to the forum, because I could not find any related threads.
I am using the Metricbeat to monitor different metrics concerning cluster health. One of which is the CPU usage for which I'm using the system.load.norm.1 field.
Now I am wondering why values associated with this field still exceed 100% from time to time, if it clearly should not be the case for a normalized value.
Google searches as well as the Kibana docs did not present an answer to my question (at least I could not find any) so maybe there are some people around here who could clarify this behavior.
Those metric values are basically just the values from running top. The metricbeat docs explain that field as such:
system.load.norm.1
type: scaled_float
Load for the last minute divided by the number of cores.
One interesting thing about top values is that they can, and regularly do, exceed 100%. This is explained on this stackexchange post, which also mentions "Solaris mode", which sounds like what that field from metricbeat is showing.
So if you're seeing a value of 100%, and you have 4 cores, it means you have a load over 4.0 ((load * 100) / # of cores). You should be able to confirm that by looking at the value for system.load.1, the non-normalized "load average for the last minute" value.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.