Metricbeat 8.x – Why were os.cpu.load_average.5m and 15m removed?

Elastic Observability Metrics
1

We observed a behavioral change between Metricbeat 7.17 and 8.x (tested with 8.13) regarding Elasticsearch node OS CPU load averages.

In Metricbeat 7.17, metricset exported:

  • node_stats.os.cpu.load_average.1m

  • node_stats.os.cpu.load_average.5m

  • node_stats.os.cpu.load_average.15m

In Metricbeat 8.x, only:

  • node_stats.os.cpu.load_average.1m

is exported, even though the Elasticsearch Nodes Stats API (/_nodes/stats/os) still returns 1m, 5m, and 15m.

I couldn’t find:

  • A breaking change note

  • A deprecation notice

  • Or documentation explaining why 5m and 15m were removed

Questions:

  1. Was the removal of 5m and 15m load averages in Metricbeat 8.x an intentional design decision?

  2. If so, what was the rationale (e.g., Stack Monitoring usage, ECS alignment, field reduction)?

  3. Is the recommended approach to collect 5m / 15m load averages now via a custom integration?

Thanks in advance.

2

Hello @Suresh_Ghatuwa

In the documentation :

Example shows only 1m :

image
image439×274 6.6 KB

So definitely there must be changes as you reported in collection of load_avg

Thanks!!

1 Like