nfv/nfv/nfv-debug-tools/histogram_analysis/plotter.py

492 lines
24 KiB
Python
Executable File

#!/usr/bin/env python
################################################################################
#
# Copyright (c) 2017 Wind River Systems, Inc.
#
# SPDX-License-Identifier: Apache-2.0
#
################################################################################
#
# Description: Graphs nfv-vim histogram data from CSV files to html page and saves
# saves it locally.
#
# Behaviour : The script takes in arguments from the command line such as specific
# process names, or the name of a grouping of processes, and graphs
# them in a local html file using plotly. The x-axis displays
# datestamps corresponding to when the sample in the csv file was taken,
# and the y-axis displays either the average execution time of the
# processes during each sample, or the difference in total hits that
# process experienced from one sample period to the previous sample
# period. Both average execution times and the delta hit count can
# be displayed on the same graph using two y-axes. The CSV files must
# be generated prior to running this script by running Histogram.sh.
# A config called logplot.cfg will be generated the first time this
# script is run and will automatically populate itself with all processes
# listed in the csv/ directory. Change the N to a Y in the right column
# of the config file to have that process graphed when running this script
# via config settings. Groupings of processes can also be made under the
# groups section by following the same N/Y format as above. When a group
# name is specified all processes listed under that group name will be
# graphed if they have a Y in their right-column.
#
# To run this script ensure that plotly is installed.
# To do this enter: sudo pip install plotly
#
#
# If no arguments are entered when running this script it will default to running
# the proceses in logplot.cfg with a Y in their rightmost column, and will display
# the average execution time on the y-axis for all available samples.
#
################################################################################
import os
import csv
import sys
import time
import plotly
import plotly.graph_objs as go
from plotly.graph_objs import Scatter, Layout
from plotly import tools
from glob import iglob
import subprocess
from builtins import input
dir = os.path.dirname(__file__)
fig = plotly.graph_objs.graph_objs.Figure
pth = os.path.join(dir, 'csv/')
execTime = False # Indicates if average execution time is to be graphed or not
default = False # Indicates no commands were entered and to run with default settings (run config with -t option)
oneAxis = False # Causes the generated graph to have two y-axes sharing an x-axis with both avg execution time and hits being graphed
config = False # Indicates whether to pull process names from logplot.cfg or not
hits = False # Indicates if the delta of hits between samples is to be graphed
markers = False
lines = False
timestamp = []
dateRange = []
warnings = []
procs = []
group = []
graphName = ""
plotType = ""
def helpMessage():
print("\n" + "-" * 120)
print("NFV-VIM Histogram Graphing Script\n")
print("This script is meant to graph average execution times and the delta of hits between sample periods for processes in nfv-vim logs.\n")
print("Usage:\n")
print(" -c ... runs from the logplot.cfg (c)onfig file. All processes in the first list with a Y\n"
" in the far-right column will be included in the generated graph.\n")
print(" -d ... command used to specify a (d)ate range within which you would like to see log data.\n"
" The format is YYYY/MM/DD-YYYY/MM/DD with the lower bound on the left, and the upper\n"
" bound on the right. The range is up to and including the bounds. To have a bound simply\n"
" cover all datestamps before or after a bound, omit the undefined bound. Only one bound can be\n"
" unspecified in this way.\n"
" e.g. -d 2016/12/01-2016/12/12\n"
" -d -2016/12/12 To use all logs prior to and including 2016/12/12\n"
" -d 2016/12/01- To use all logs after and including 2016/12/01\n")
print(" -t ... used to indicate that you would like the graph to display average execution (t)imes\n"
" along the y-axis.\n")
print(" -h ... used to indicate that you would like the graph to display the dela of (h)its between\n"
" sample periods on the y-axis.\n")
print(" -l ... used to set the graph to be a line graph. (Can be used with -m as well)\n")
print(" -m ... used to set the graph to be a scatterplot. (Can be used with -l as well)\n")
print(" -lm ... used to set the graph to be a scatterplot with connecting lines. This same effect can also be\n"
" achieved by using -l -m\n")
print(" -n ... used to (n)ame the file that will be generated by this script. Files can be found in the\n"
" Graphs/ directory, found inside the directory containing this script. Do not include spaces in\n"
" the file name. If no name is specified, the name will default to the timestamp from when the\n"
" script was run.\n"
" e.g. 01-24-2017.html\n")
print(" -oneaxis ... used to generate a graph with two Y-axes sharing an x-axis. Average execution time's y-axis is\n"
" on the right, and delta Hits per sample's y-axis is on the left. Used to look for correlations. The \n"
" name of the process being graphed will have _time or _hits appended to it so you can tell which\n"
" y-axis to relate it to. Only works if both -h and -t flags are used. Can be used for multiple processes.\n"
" e.g. -h -t -oneaxis --p process1 process2\n")
print(" --g ... will run the script for processes specified in logplot.cfg under the (G)roups heading.\n"
" All processes listed under the named group's heading will be included in the graph.\n"
" Space-delimit the groups to be included. This must be the last command entered.\n"
" e.g. --g group1 group2\n")
print(" --p ... follow this with a space-delimited list of (p)rocesses you would like to graph together.\n"
" This must be the last command entered.\n")
print(" --update ... This will update the master list of process at the beginning of the logplot.cfg file:\n"
" Processes not currently listed in the master list will be added and their run status set to N.\n\n")
print("Note: If neither the -t nor -h tag is used, the script will default to display the average execution time on the y-axis.\n\n")
print("Examples:\n")
print("./plotter.py -c -d 2016/12/3-2016/12/10 -t -n ConfigOutput_Dec_3-10 ... This will graph all processes with a Y in their\n"
" right-most column in the config file, using logs\n"
" with a timestamp between Dec 3rd and 10th 2016,\n"
" and will display their average execution time in\n"
" the y-axis. The file will be called\n"
" ConfigOutput_Dec_3-10.html")
print("./plotter.py -h -t --g group1 ... This will generate two graphs, one with the delta of hits\n"
" on the y-axis, and the other with the average execution time\n"
" in the y-axis, for processes listed under group1 in\n"
" logplot.cfg.\n"
" period will be displayed on the y-axis.\n")
print("./plotter.py ... This will run the default settings, which are to run\n"
" for the processes enabled in the master list in\n"
" the config file, to use log information for all dates\n"
" available, to show average execution time on the y-axis,\n"
" and to name the file with the current day's datestamp.")
print("-" * 120)
# Appends new processes found via CSV filenames to the master process list in logplot.cfg if there are not already present.
# If logplot.cfg has not been generated yet, this will create it and add process names found in filenames in ./csv
def updater(configExists=True):
procs = []
existingProcs = []
newProcs = []
position = 0 # Tracks position of the end of the master process list so new processes can be added above it.
os.chdir(pth)
for name in iglob("*.csv"):
procs.append(str(name)[:-4])
os.chdir("..")
if not configExists:
f = open(os.path.join(dir, 'logplot.cfg'), "w")
for p in procs:
f.write(p + " " * (59 - len(p)) + "N\n")
f.write("#" * 20 + "END OF PROCESS LIST" + "#" * 21 + "\n\n")
f.write("#" * 27 + "GROUPS" + "#" * 27 + "\n")
f.write("#GroupSTART\n")
f.write("GroupName=ExampleGroupName1\n")
f.write("ExampleProcessName1" + " " * 40 + "N\n")
f.write("ExampleProcessName2" + " " * 40 + "N\n")
f.write("#GroupEND\n")
f.write("-" * 60 + "\n")
f.write("GroupName=ExampleGroupName2\n")
f.write("ExampleProcessName3" + " " * 40 + "N\n")
f.write("ExampleProcessName4" + " " * 40 + "N\n")
f.write("#GroupEND\n")
f.write("#" * 20 + "END OF GROUPS" + "#" * 27)
f.close()
else:
with open(os.path.join(dir, 'logplot.cfg'), "r+") as f:
cfgLines = f.read().splitlines()
for cfgProc in cfgLines:
if "#END" in cfgProc:
break
existingProcs.append(cfgProc.split()[0])
position += 1
for p in procs:
if p not in existingProcs:
newProcs.append(p + " " * (59 - len(p)) + "N")
procs = cfgLines[:position] + newProcs + cfgLines[position:]
f.seek(0)
f.write("\n".join(procs))
f.truncate()
f.close()
# Appends process names found in the specified group to the list of processes to be graphed.
def gCommand(groups):
procs = []
f = open(os.path.join(dir, 'logplot.cfg'), "r")
cfgLines = f.read().splitlines()
for g in groups:
groupFound = False
finishedGroup = False
for i in range(len(cfgLines)):
liNum = i
if str("GroupName=" + g) == cfgLines[i].strip():
groupFound = True
while not finishedGroup:
liNum += 1
if "GroupEND" in cfgLines[liNum]:
finishedGroup = True
else:
cfgLine = cfgLines[liNum].split()
if cfgLine[1] == "Y":
procs.append(cfgLine[0])
else:
break
else:
if not groupFound:
warnings.append("WARNING: The following group could not be found: %s\n\t\t Please check your logplot.cfg file for the intended group name." % (g,))
f.close()
return procs
# Appends processes explicitly named by the user to the list of processes to be run.
# If the process name specified using the --p command does not match the name of any processes taken from .csv filenames, the user is given
# a list of known processes containing the name they entered. If they enter one of the provided names, it will be added to the list. If the
# user enters "s", the process in question will be skipped and the script will continue. If they user enters "q" the script will exit.
def pCommand(pList):
procList = []
for i in range(len(pList)):
csvFile = str(pList[i]) + ".csv"
procName = str(pList[i])
isFile = False
if os.path.isfile(os.path.join(pth, csvFile)):
isFile = True
procList.append(pList[i])
else:
while(not isFile):
print("\nFiles containing keyword: %s" % (str(procName)))
csvFile = str(procName) + ".csv"
for root, directories, filenames in os.walk(pth):
for filename in filenames:
if procName.lower() in filename.lower():
if (str(procName) + ".csv") == str(filename):
isFile = True
procList.append(str(procName).strip())
break
else:
print(" " + filename[:-4])
else:
procName = str(input("\nEnter the corrected process name, q to quit, or s to skip: ")).strip()
if procName == "s":
isFile = True
break
elif procName == "q":
sys.exit()
return procList
# Stores the average execution time, or delta hit count data into into a plotly graph obj, and restricts sample to be within a certain
# date range if specified. If plots is 1, one graph will be generated. If plots is 2, two graphs will be generated with one above the other.
def storeGraphData(procs, dateRange=[], execTime=False, hits=False, plots=1):
graphData = {}
prevHitTotal = 0
timeList = [[] for p in range(len(procs))]
dateList = [[] for p in range(len(procs))]
hitList = [[] for p in range(len(procs))]
if dateRange:
for i in range(len(procs)):
csvFile = str(procs[i]) + ".csv"
with open(os.path.join(pth, csvFile), 'rb') as f:
reader = csv.reader(f, delimiter=',', quoting=csv.QUOTE_NONE)
for ts, at, h, n in reader:
t = ts.split("T")
date = ''.join(x for x in t[0].split('-'))
if (int(date) >= int(dateRange[0])) and (int(date) <= int(dateRange[1])):
timeList[i].append(at)
dateList[i].append(str(ts[0:10:1] + " " + ts[11:]))
hitList[i].append(int(h) - prevHitTotal)
prevHitTotal = int(h)
f.close()
hitList[i][0] = None
graphData['trace' + str(i)] = go.Scatter(x=dateList[i],
y=timeList[i] if execTime else hitList[i],
mode=plotType,
name=(procs[i] if not oneAxis else (procs[i] + "_" + ("time" if execTime else "hits"))))
if plots == 1:
fig.append_trace(graphData['trace' + str(i)], 1, 1)
elif plots == 2:
fig.append_trace(graphData['trace' + str(i)], 2, 1)
else:
for i in range(len(procs)):
csvFile = str(procs[i]) + ".csv"
with open(os.path.join(pth, csvFile), 'rb') as f:
reader = csv.reader(f, delimiter=',', quoting=csv.QUOTE_NONE)
for ts, at, h, n in reader:
timeList[i].append(at)
dateList[i].append(str(ts[0:10:1] + " " + ts[11:]))
hitList[i].append(int(h) - prevHitTotal)
prevHitTotal = int(h)
f.close()
hitList[i][0] = None
graphData['trace' + str(i)] = go.Scatter(x=dateList[i],
y=timeList[i] if execTime else hitList[i],
mode=plotType,
name=(procs[i] if not oneAxis else (procs[i] + "_" + ("time" if execTime else "hits"))))
if plots == 1:
fig.append_trace(graphData['trace' + str(i)], 1, 1)
elif plots == 2:
fig.append_trace(graphData['trace' + str(i)], 2, 1)
# Formats the graph by adding axis titles, changing font sizes, setting there to be two separate graphs or two graphs sharing an x-axis etc.
def formatGraph(two, oneAxis):
fig['layout'].update(showlegend=True)
if two:
if oneAxis:
fig['layout']['xaxis1'].update(title='Timestamp', titlefont=dict(size=20, color='#4d4d4d'))
fig['layout']['yaxis1'].update(title='Hits Per Sample', titlefont=dict(size=20, color='#4d4d4d'))
fig['layout']['yaxis2'].update(title='Average Execution Time (milliseconds)', anchor='x', overlaying='y', side='right', position=1, titlefont=dict(size=20, color='#4d4d4d'))
else:
fig['layout']['xaxis1'].update(title='Timestamp', titlefont=dict(size=20, color='#4d4d4d'))
fig['layout']['yaxis1'].update(title='Average Execution Time (milliseconds)', titlefont=dict(size=20, color='#4d4d4d'))
fig['layout']['xaxis2'].update(title='Timestamp', titlefont=dict(size=20, color='#4d4d4d'))
fig['layout']['yaxis2'].update(title='Hits Per Sample', titlefont=dict(size=20, color='#4d4d4d'))
fig['layout'].update(title=graphName, titlefont=dict(size=26))
else:
fig['layout'].update(
title=graphName,
xaxis=dict(
title="Timestamp",
titlefont=dict(
family='Courier New, monospace',
size=18,
color='#4d4d4d'
)
),
yaxis=dict(
title="Average Execution Time (milliseconds)" if execTime else "Hits Per Sample",
titlefont=dict(
family='Courier New, monospace',
size=18,
color='#4d4d4d'
)
)
)
# Sets the name of the saved html file.
def setFilename(graphName):
validName = False
if not os.path.exists("Graphs/"):
os.makedirs("Graphs/")
os.chdir(os.path.join(dir, 'Graphs/'))
if not graphName:
graphName = time.strftime("%m-%d-%Y")
if os.path.exists(str(graphName + ".html")):
n = 1
while(not validName):
if os.path.exists(str(graphName + "(" + str(n) + ").html")):
n += 1
else:
graphName = graphName + "(" + str(n) + ")"
validName = True
return graphName
print("Welcome to plotter, type --help for information")
# Checks that plotly is installed, otherwise graphs cannot be generated.
plotCheck = subprocess.getstatusoutput("pip list | grep plotly")
if plotCheck[0] == 0:
if "plotly" not in plotCheck[1]:
print("\n\tWARNING: Plotly is not installed on your system.\n\tPlease install it with: sudo pip install plotly\n")
sys.exit()
# Checks to see if logplot.cfg already exists, creates it if not.
if not os.path.isfile(os.path.join(dir, 'logplot.cfg')):
print("Generating logplot.cfg")
updater(False)
print("logplot.cfg created.")
if not os.path.isdir('./csv'):
print("\n\tWARNING: ./csv directory is missing. Please run Histogram.sh or make sure directory has not been renamed.\n")
sys.exit()
command = sys.argv # Takes arguments from the command line
if len(command) == 1:
print("Running with default settings.")
default = True
else:
for i in range(1, len(command)):
if command[i] == "-c": # Use config file
config = True
elif command[i] == "--g": # Groups
for j in range(i + 1, len(command)):
group.append(command[j])
procs = gCommand(group)
break
elif command[i] == "-t": # Average execution time
execTime = True
elif command[i] == "-h": # Delta hits between samples
hits = True
elif command[i] == "-l": # Graph with lines
lines = True
elif command[i] == "-m": # Graph with markers (scatter)
markers = True
elif command[i] == "-lm": # Graph with lines and markers
lines = True
markers = True
elif command[i] == "-d": # Date range
dateRange = command[i + 1].split('-')
if dateRange[0]:
lower = dateRange[0].split("/")
dateRange[0] = lower[0] + lower[1].zfill(2) + lower[2].zfill(2)
else:
dateRange[0] = "0" * 8
if dateRange[1]:
upper = dateRange[1].split("/")
dateRange[1] = upper[0] + upper[1].zfill(2) + upper[2].zfill(2)
else:
dateRange[1] = "9" * 8
i += 1
elif command[i] == "-n": # Name of file to be generated
graphName = command[i + 1]
i += 1
elif command[i] == "-oneaxis": # Have hit and time data displayed on same graph
oneAxis = True
elif (command[i] == "--help") or (command[i] == "--h"): # Print help message and exit script
helpMessage()
sys.exit()
elif command[i] == "--p": # User-specified processes
for j in range(i + 1, len(command)):
procs.append(command[j])
procs = pCommand(procs)
break
elif command[i] == "--update":
print("Updating...")
updater()
print("Update complete.")
sys.exit()
# If neither average execution time nor delta hit count are specified to be shown, default to showing average execution time.
if (not execTime) and (not hits):
execTime = True
# Default settings can be changed as desired.
if default:
config = True
execTime = True
if (lines and markers):
plotType = "lines+markers"
elif lines:
plotType = "lines"
else:
plotType = "markers"
if config:
f = open(os.path.join(dir, 'logplot.cfg'), "r")
procList = f.read().splitlines()
for p in procList:
if "#END" in p:
break
cfgLine = p.split()
if cfgLine[1] == "Y":
csvFile = cfgLine[0] + ".csv"
if os.path.exists(os.path.join(pth, csvFile)):
procs.append(cfgLine[0])
else:
warnings.append("WARNING: %s does not exist." % (csvFile,))
f.close()
# If both average execution time and delta hits are specified to be shown, generate two graphs if -oneaxis wasn't specified.
# If only one of execution time and delta hits was specified, generate one graph.
if procs:
if (execTime and hits):
if(not oneAxis):
fig = tools.make_subplots(rows=2, cols=1)
storeGraphData(procs, dateRange, execTime, False, 1)
storeGraphData(procs, dateRange, False, hits, 2)
else:
fig = tools.make_subplots(rows=1, cols=1)
storeGraphData(procs, dateRange, False, hits, 1)
storeGraphData(procs, dateRange, execTime, False, 1)
else:
fig = tools.make_subplots(rows=1, cols=1)
storeGraphData(procs, dateRange, execTime, hits)
formatGraph((execTime and hits), oneAxis)
# Generates the plot
plotly.offline.plot(fig, filename=setFilename(graphName) + ".html")
else:
warnings.append("NO GRAPH GENERATED BECAUSE NO VALID GROUP OR PROCESS NAME SPECIFIED.")
# If any warnings occured, print them
if warnings:
print("\n\t" + ("\n\t").join(warnings) + "\n")