2

In Elasticsearch node stat API when I send a query for OS stat:

curl -XGET "http://esls1.ping-service.com:9200/_nodes/stats/os"

In the response I get a metric load_average:

"load_average": [0,0.04,0.13]

What does is means?

Tariq
  • 2,489
  • 11
  • 36
  • 61

2 Answers2

2

That is the currently calculated average load of the system and how this is obtained is specific to the operating system Elasticsearch is installed on.

ES uses Sigar to get this kind of information. The three numbers represent average loads calculated for 1 minute, 5 minutes and 15 minutes intervals.

For linux, for example, Sigar uses /proc/loadavg to get this information from the system. You can find more about this specific calculation in this SO post.

For AIX, Sigar is using perfstat_cpu_total subroutine, if I'm not mistaken to get the same information.

Community
  • 1
  • 1
Andrei Stefan
  • 51,654
  • 6
  • 98
  • 89
  • Sigar is not used in Elasticsearch anymore since the first Beta of 2.0.0: https://github.com/elastic/elasticsearch/pull/12010 https://github.com/elastic/elasticsearch/issues/11034 – Wolli Feb 18 '20 at 11:44
0

Sigar is not used in Elasticsearch anymore since the first Beta of 2.0.0: github.com/elastic/elasticsearch/pull/12010 github.com/elastic/elasticsearch/issues/11034

since then, they switched to generic OS load metrics. similar to what you see with the top command. see here for an explanation what this means: https://askubuntu.com/questions/532845/what-is-system-load

beware: this means if you run ES in a docker container, the load shown will be actually from the host machine, and not from only the docker container!

Wolli
  • 748
  • 1
  • 8
  • 12