1#!/usr/bin/python
2
3import matplotlib
4import matplotlib.pyplot as plt
5from matplotlib import gridspec
6import numpy as np
7import datetime
8import json
9import pandas
10import argparse
11import os
12matplotlib.style.use('ggplot')
13
14# sacrificial plot for single legend
15matplotlib.rcParams['figure.figsize'] = 1, 1
16randFig = plt.figure()
17randAx = plt.subplot()
18randAx.plot(0, 0, color='green', label='good', marker='+')
19randAx.plot(0, 0, color='red', label='failed', marker='x')
20randAx.plot(0, 0, color='black', label='sent', linestyle='--')
21randAx.plot(0, 0, color='green', label='50th quantile')
22randAx.plot(0, 0, color='orange', label='90th quantile')
23randAx.plot(0, 0, color='red', label='99th quantile')
24handles, labels = randAx.get_legend_handles_labels()
25
26# big ol' plotting method
27def plot_section(all_data, title, outputPath):
28 # group calls by the endpoint/method
29 actions = all_data.groupby('action')
30 h = len(actions.groups.keys())
31 matplotlib.rcParams['figure.figsize'] = 20, 3 * h
32
33 fig = plt.figure()
34 fig.legend(handles, labels, ncol=6, fontsize=16, framealpha=0, loc='upper center')
35 if title is not None:
36 fig.suptitle(title, fontsize=20, y=0.93)
37 gs = gridspec.GridSpec(h, 3)
38
39 # figure out left and right datetime bounds
40 started = all_data['sent'].min()
41 stopped = all_data['finished'].max()
42
43 i = 0
44 # plot one row of charts for each endpoint/method combination
45 for section in actions.groups.keys():
46 # setup the tree charts
47 ax = fig.add_subplot(gs[i, 0])
48 ax.set_title(section)
49 ax.set_xlim(started, stopped)
50 ax2 = fig.add_subplot(gs[i, 2])
51 ax2.set_xlim(started, stopped)
52 ax3 = fig.add_subplot(gs[i, 1])
53 ax3.set_xlim(started, stopped)
54
55 # find the maximum y value and set it across all three charts
56 calls = actions.get_group(section)
57 tookMax = calls['took'].max()
58 ax.set_ylim(0, tookMax+tookMax*0.1)
59 ax2.set_ylim(0, tookMax+tookMax*0.1)
60 ax3.set_ylim(0, tookMax+tookMax*0.1)
61
62 groups = calls.groupby('type')
63 if groups.groups.get('error', False) is not False:
64 bad = groups.get_group('error')
65 ax.plot_date(bad['finished'], bad['took'], color='red', marker='x', label='error')
66
67 bad_rate = bad.set_index('finished')
68 bad_rate['rate'] = [0] * len(bad_rate.index)
69 bad_rate = bad_rate.resample('5S').count()
70 bad_rate['rate'] = bad_rate['rate'].divide(5)
71 rateMax = bad_rate['rate'].max()
72 ax2.plot_date(bad_rate.index, bad_rate['rate'], linestyle='-', marker='', color='red', label='error')
73 if groups.groups.get('good', False) is not False:
74 good = groups.get_group('good')
75 ax.plot_date(good['finished'], good['took'], color='green', marker='+', label='good')
76
77 good_rate = good.set_index('finished')
78 good_rate['rate'] = [0] * len(good_rate.index)
79 good_rate = good_rate.resample('5S').count()
80 good_rate['rate'] = good_rate['rate'].divide(5)
81 rateMax = good_rate['rate'].max()
82 ax2.plot_date(good_rate.index, good_rate['rate'], linestyle='-', marker='', color='green', label='good')
83 ax.set_ylabel('Latency (ms)')
84
85 # calculate the request rate
86 sent_rate = pandas.DataFrame(calls['sent'])
87 sent_rate = sent_rate.set_index('sent')
88 sent_rate['rate'] = [0] * len(sent_rate.index)
89 sent_rate = sent_rate.resample('5S').count()
90 sent_rate['rate'] = sent_rate['rate'].divide(5)
91 if sent_rate['rate'].max() > rateMax:
92 rateMax = sent_rate['rate'].max()
93 ax2.plot_date(sent_rate.index, sent_rate['rate'], linestyle='--', marker='', color='black', label='sent')
94 ax2.set_ylim(0, rateMax+rateMax*0.1)
95 ax2.set_ylabel('Rate (per second)')
96
97 # calculate and plot latency quantiles
98 calls = calls.set_index('finished')
99 calls = calls.sort_index()
100 quan = pandas.DataFrame(calls['took'])
101 for q, c in [[.5, 'green'], [.9, 'orange'], [.99, 'red']]:
102 quanN = quan.rolling(500, center=True).quantile(q)
103 ax3.plot(quanN['took'].index, quanN['took'], color=c)
104
105 ax3.set_ylabel('Latency quantiles (ms)')
106
107 i += 1
108
109 # format x axes
110 for ax in fig.axes:
111 matplotlib.pyplot.sca(ax)
112 plt.xticks(rotation=30, ha='right')
113 majorFormatter = matplotlib.dates.DateFormatter('%H:%M:%S')
114 ax.xaxis.set_major_formatter(majorFormatter)
115
116 # save image
117 gs.update(wspace=0.275, hspace=0.5)
118 fig.savefig(outputPath, bbox_inches='tight')
119
120# and the main event
121parser = argparse.ArgumentParser()
122parser.add_argument('chartData', type=str, help='Path to file containing JSON chart output from load-generator')
123parser.add_argument('--output', type=str, help='Path to save output to', default='latency-chart.png')
124parser.add_argument('--title', type=str, help='Chart title')
125args = parser.parse_args()
126
127with open(args.chartData) as data_file:
128 stuff = []
129 for l in data_file.readlines():
130 stuff.append(json.loads(l))
131
132df = pandas.DataFrame(stuff)
133df['finished'] = pandas.to_datetime(df['finished']).astype(datetime.datetime)
134df['sent'] = pandas.to_datetime(df['sent']).astype(datetime.datetime)
135df['took'] = df['took'].divide(1000000)
136
137plot_section(df, args.title, args.output)
View as plain text