Elastic APM Python - System Metrics don't show process related metrics like memory on Kibana

Elastic Observability APM
1

I have a standalone Python application.

The python process is not using any framework. And is a simple standalone python process.

This has been successfully integrated with Elastic APM (+ElasticSearch +Kibana) for capturing telemetries.



Python version: 3.7

elastic-apm python agent: 5.8.0

Elastic Search, APM and Kibana Version: 7.7.1



As mentioned in the official doc, I have used the following statements to start capturing metrics from my python process

client = elasticapm.Client(service_name="MY_SERVICE",service_node_name="MY_SERVICE_NODE",server_url="=http:///apm-server:8200",recording="true",environment="PROD")
elasticapm.instrument()

But on Kibana, I am able to see only the following 3 system metrics (under 2 visualizations):

  1. Host's CPU Usage in percentage
  2. Process specific CPU Usage in percentage
  3. Host's Memory Usage in percentage

As per the python code analysis, as well as per what I have read. Elastic APM Agent collects other process related metrics like:

  1. Memory used by Process
  2. Process uptime

These I am unable to see on the Kibana - APM - Metrics Visualization

Refer the screenshot
Refer the screenshot





Additionally, I expect the Elastic APM Python agent to collect other informations like:

  1. Process's thread level CPU Usage
  2. Number of threads in process
  3. Garbage collection stats
  4. IO Stats

Which are already available for Elastic APM Java agent.


Refer the screenshot

Refer the screenshot




Need help in understanding, whether these process related metrics are available in Kibana - APM or not?

If available, how to enable the visualization?

2

Hello @Debashish_BHARALI!

You can see the list of metrics that the python agent collects here: https://www.elastic.co/guide/en/apm/agent/python/current/metrics.html

Unfortunately, the metrics we collect are not as diverse as those collected by the java agent. Some of this is because the java agent can take advantage of the JVM for these metrics, but some of this is because it hasn't been at the top of our priority list.

You're welcome to open a feature request for specific metrics you'd like to see in the future.

1 Like
3

Hi @basepi

Thanks for your response.

What I can see from the reference you shared, the python agent collects the following:
system.process.memory.size
Which is the total virtual memory of the process.

I was intrigued by that, but was not able to find the corresponding visualization. Is the corresponding available?

I have multiple industry common metrics in mind which should be critical for performance monitoring.
I would certainly raise the corresponding feature requests. Thanks for sharing the URL.

4

Not all metrics have built-in visualizations. For example, most users don't care what the actual memory size of their app is -- rather, they care if the host is running out of memory. Thus, system memory usage is one of the default metric visualizations.

However, the data are still there, ready to be used to make your own visualizations and dashboards. Try searching

system.process.memory.size : * and agent.name : "python"

in kibana on the apm-* index pattern to see a recent sampling for the python agent.

5

Hi @basepi
Thanks for the response. Successfully got it.

6

This topic was automatically closed 20 days after the last reply. New replies are no longer allowed.