forked from kubernetes/test-infra
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathanalyze-memory-profiles.py
executable file
·161 lines (145 loc) · 5.99 KB
/
analyze-memory-profiles.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
#!/usr/bin/env python3
# Copyright 2021 The Kubernetes Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# This script is meant to be used to analyze memory profiles created by the Prow binaries when
# the --profile-memory-usage flag is passed. The interval of profiling can be set with the
# --memory-profile-interval flag. This tool can also be used on the output of the sidecar utility
# when the sidecar.Options.WriteMemoryProfile option has been set. The tools will write sequential
# profiles into a directory, from which this script can load the data, create time series and
# visualize them.
import os
import pathlib
import subprocess
import sys
from datetime import datetime
import matplotlib.dates as mdates
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from matplotlib.font_manager import FontProperties
if len(sys.argv) != 2:
print("[ERROR] Expected the directory containing profiles as the only argument.")
print("Usage: {} ./path/to/profiles/".format(sys.argv[0]))
sys.exit(1)
profile_dir = sys.argv[1]
def parse_bytes(value):
# we will either see a raw number or one with a suffix
value = value.decode("utf-8")
if not value.endswith("B"):
return float(value)
suffix = value[-2:]
multiple = 1
if suffix == "KB":
multiple = 1024
elif suffix == "MB":
multiple = 1024 * 1024
elif suffix == "GB":
multiple = 1024 * 1024 * 1024
return float(value[:-2]) * multiple
overall_name = "overall".encode("utf-8")
dates_by_name = {overall_name: []}
flat_usage_over_time = {overall_name: []}
cumulative_usage_over_time = {overall_name: []}
max_usage = 0
for subdir, dirs, files in os.walk(profile_dir):
for file in files:
full_path = os.path.join(subdir, file)
date = datetime.fromtimestamp(pathlib.Path(full_path).stat().st_mtime)
output = subprocess.run(
["go", "tool", "pprof", "-top", "-inuse_space", full_path],
check=True, stdout=subprocess.PIPE
)
# The output of go tool pprof will look like:
#
# File: sidecar
# Type: inuse_space
# Time: Mar 19, 2021 at 10:30am (PDT)
# Showing nodes accounting for 66.05MB, 100% of 66.05MB total
# flat flat% sum% cum cum%
# 64MB 96.90% 96.90% 64MB 96.90% google.golang.org/api/internal/gensupport...
#
# We want to parse all of the lines after the header and metadata.
lines = output.stdout.splitlines()
usage = parse_bytes(lines[3].split()[-2])
if usage > max_usage:
max_usage = usage
data_index = 0
for i in range(len(lines)):
if lines[i].split()[0].decode("utf-8") == "flat":
data_index = i + 1
break
flat_overall = 0
cumulative_overall = 0
for line in lines[data_index:]:
parts = line.split()
name = parts[5]
if name not in dates_by_name:
dates_by_name[name] = []
dates_by_name[name].append(date)
if name not in flat_usage_over_time:
flat_usage_over_time[name] = []
flat_usage = parse_bytes(parts[0])
flat_usage_over_time[name].append(flat_usage)
flat_overall += flat_usage
if name not in cumulative_usage_over_time:
cumulative_usage_over_time[name] = []
cumulative_usage = parse_bytes(parts[3])
cumulative_usage_over_time[name].append(cumulative_usage)
cumulative_overall += cumulative_usage
dates_by_name[overall_name].append(date)
flat_usage_over_time[overall_name].append(flat_overall)
cumulative_usage_over_time[overall_name].append(cumulative_overall)
plt.rcParams.update({'font.size': 22})
fig = plt.figure(figsize=(30, 18))
plt.subplots_adjust(right=0.7)
ax = plt.subplot(211)
for name in dates_by_name:
dates = mdates.date2num(dates_by_name[name])
values = flat_usage_over_time[name]
# we only want to show the top couple callsites, or our legend gets noisy
if max(values) > 0.01 * max_usage:
ax.plot_date(dates, values,
label="{} (max: {:,.0f}MB)".format(name.decode("utf-8"), max(values) / (1024 * 1024)),
linestyle='solid')
else:
ax.plot_date(dates, values, linestyle='solid')
ax.set_yscale('log')
ax.set_ylim(bottom=10*1024*1024)
formatter = ticker.FuncFormatter(lambda y, pos: '{:,.0f}'.format(y / (1024 * 1024)) + 'MB')
ax.yaxis.set_major_formatter(formatter)
plt.xlabel("Time")
plt.ylabel("Flat Space In Use (bytes)")
plt.title("Space In Use By Callsite")
fontP = FontProperties()
fontP.set_size('xx-small')
plt.legend(bbox_to_anchor=(1, 1), loc='upper left', prop=fontP)
ax = plt.subplot(212)
for name in dates_by_name:
dates = mdates.date2num(dates_by_name[name])
values = cumulative_usage_over_time[name]
# we only want to show the top couple callsites, or our legend gets noisy
if max(values) > 0.01 * max_usage:
ax.plot_date(dates, values,
label="{} (max: {:,.0f}MB)".format(name.decode("utf-8"), max(values) / (1024 * 1024)),
linestyle='solid')
else:
ax.plot_date(dates, values, linestyle='solid')
ax.set_yscale('log')
ax.set_ylim(bottom=10*1024*1024)
ax.yaxis.set_major_formatter(formatter)
plt.xlabel("Time")
plt.ylabel("Cumulative Space In Use (bytes)")
fontP = FontProperties()
fontP.set_size('xx-small')
plt.legend(bbox_to_anchor=(1, 1), loc='upper left', prop=fontP)
plt.show()