plot-docker-stats.py
ActivePublic
Actions

Authored by JMeybohm on May 7 2021, 1:05 PM.

Tags

None

Referenced Files

	F34449190: plot-docker-stats.py
	May 10 2021, 2:25 PM

	F34445119: plot-docker-stats.py
	May 7 2021, 1:05 PM

Subscribers

None

	#!/bin/env python3

	import argparse
	import json

	import pandas as pd
	import plotly.graph_objects as go
	from plotly.subplots import make_subplots


	def calculate_CPU_usage(record):
	"""Return the used CPU delta (in ms) and CPU usage in percent since last tick"""
	cpu_percent = 0.0
	cpu_delta = (
	record["cpu_stats"]["cpu_usage"]["total_usage"]
	- record["precpu_stats"]["cpu_usage"]["total_usage"]
	)
	system_delta = (
	record["cpu_stats"]["system_cpu_usage"]
	- record["precpu_stats"]["system_cpu_usage"]
	)
	cpu_count = record["cpu_stats"]["online_cpus"]
	if system_delta > 0 and cpu_delta > 0:
	cpu_percent = (cpu_delta / system_delta) * cpu_count * 100
	if cpu_delta > 0:
	cpu_delta = cpu_delta / 1000000

	return (cpu_delta, cpu_percent)


	def read_json(fileobj):
	j = []
	first_line = True
	for record in fileobj:
	# Skip the first line it does not have previous recordings
	if first_line:
	first_line = False
	continue
	try:
	record = json.loads(record)
	except json.decoder.JSONDecodeError:
	# probably last line not completely flushed
	continue
	cpu_delta, cpu_percent = calculate_CPU_usage(record)

	tick = dict(read=record["read"])
	tick["cpu_delta"] = cpu_delta
	tick["cpu_percent"] = cpu_percent
	# Calculate the time (in ms) the container was throttled since last tick
	tick["cpu_throttled_ms"] = (
	record["cpu_stats"]["throttling_data"]["throttled_time"]
	- record["precpu_stats"]["throttling_data"]["throttled_time"]
	) / 1000000

	j.append(tick)
	return j


	def plot(df, title="foo", output=None):
	fig = make_subplots(
	rows=2, cols=1, shared_xaxes=True, subplot_titles=("CPU %", "CPU (trottled) ms")
	)
	fig.add_trace(
	go.Scatter(x=df["read"], y=df["cpu_percent"], name="CPU %"),
	row=1,
	col=1,
	)
	fig.add_trace(
	go.Scatter(x=df["read"], y=df["cpu_delta"], name="CPU ms"),
	row=2,
	col=1,
	)
	fig.add_trace(
	go.Scatter(x=df["read"], y=df["cpu_throttled_ms"], name="throttled ms"),
	row=2,
	col=1,
	)

	fig.update_xaxes(title_text="Time")
	fig.update_yaxes(title_text="%", range=[0, 100], row=1, col=1)
	fig.update_yaxes(title_text="ms", row=2, col=1)
	fig.update_layout(title_text=title)

	if output is None:
	fig.show()
	else:
	fig.write_html(output)


	if __name__ == "__main__":
	"""
	Generate the input file with something like:
	curl -s --unix-socket /var/run/docker.sock \
	"http://localhost/containers/${CONTAINER_ID}/stats?stream=1" \|
	jq -c '{read: .read, cpu_stats: .cpu_stats, precpu_stats: .precpu_stats}'
	"""
	parser = argparse.ArgumentParser(description="Plot docker stats CPU usage data")
	parser.add_argument("input", type=argparse.FileType("r"))
	parser.add_argument("--out", type=argparse.FileType("w"), required=False)
	args = parser.parse_args()

	j = read_json(args.input)
	df = pd.DataFrame.from_dict(j)
	plot(df, args.input.name, args.out)

Event Timeline

JMeybohm created this paste.May 7 2021, 1:05 PM

JMeybohm mentioned this in T277877: Set resource requests and limits for calico PODs.May 7 2021, 1:14 PM

JMeybohm edited the content of this paste. (Show Details)May 10 2021, 2:25 PM

plot-docker-stats.pyActivePublicActions

Event Timeline

plot-docker-stats.py
ActivePublic
Actions