[question] Does this repo only support kubernetes metrics?

[Jeffwan]

I am curious whether this custom pod autoscaler framework support fetching metrics from pod directly instead of using kubernetes metrics?

[jthomperoo]

Hi @Jeffwan, yes Custom Pod Autoscalers can fetch metrics directly from pods, see this example to see how to do that:

https://github.com/jthomperoo/custom-pod-autoscaler/tree/5d39741bd6990c5268dc46dddcac2e7e8c0067ba/example/simple-pod-metrics-python

In the metric.py script it grabs the pod IP and uses that to send a request to an arbitrary /metrics endpoint exposed by the pod:

custom-pod-autoscaler/example/simple-pod-metrics-python/metric.py

Lines 38 to 55 in 5d39741

	def metric(spec):
	# Get Pod IP
	status = spec["resource"]["status"]
	ip = status["podIP"]
	try:
	# Make request to Pod metric endpoint
	# (see ../flask-metrics/ folder for simple flask app exposing this endpoint)
	response = requests.get(f"http://{ip}:5000/metric")
	# Output whatever metrics are gathered to stdout
	sys.stdout.write(response.text)
	except HTTPError as http_err:
	# If an error occurs, output error to stderr and exit with status 1
	sys.stderr.write(f"HTTP error occurred: {http_err}")
	exit(1)
	except Exception as err:
	# If an error occurs, output error to stderr and exit with status 1
	sys.stderr.write(f"Other error occurred: {err}")
	exit(1)

You can read more about the per-pod run mode which runs metric gathering per pod in the documentation here:
https://custom-pod-autoscaler.readthedocs.io/en/latest/user-guide/metric-gathering/#per-pod

[jthomperoo]

If my response above hasn't answered your question fully just re-open this issue with a comment explaining why.
Thanks!

[Jeffwan]

@jthomperoo Thanks! I read the code and understand how it works now. kuberneresMetrics is actually additional metrics cpa can help feed into metric component, the primary focus could be per service or per pod metric scraping