| 1 | #!/usr/bin/python |
| 2 | # |
| 3 | # Copyright (C) 2013 eNovance SAS <licensing@enovance.com> |
| 4 | # Author: Erwan Velu <erwan@enovance.com> |
| 5 | # |
| 6 | # The license below covers all files distributed with fio unless otherwise |
| 7 | # noted in the file itself. |
| 8 | # |
| 9 | # This program is free software; you can redistribute it and/or modify |
| 10 | # it under the terms of the GNU General Public License version 2 as |
| 11 | # published by the Free Software Foundation. |
| 12 | # |
| 13 | # This program is distributed in the hope that it will be useful, |
| 14 | # but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | # GNU General Public License for more details. |
| 17 | # |
| 18 | # You should have received a copy of the GNU General Public License |
| 19 | # along with this program; if not, write to the Free Software |
| 20 | # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
| 21 | |
| 22 | import os |
| 23 | import fnmatch |
| 24 | import sys |
| 25 | import getopt |
| 26 | import re |
| 27 | import math |
| 28 | |
| 29 | def find_file(path, pattern): |
| 30 | fio_data_file=[] |
| 31 | # For all the local files |
| 32 | for file in os.listdir(path): |
| 33 | # If the file math the regexp |
| 34 | if fnmatch.fnmatch(file, pattern): |
| 35 | # Let's consider this file |
| 36 | fio_data_file.append(file) |
| 37 | |
| 38 | return fio_data_file |
| 39 | |
| 40 | def generate_gnuplot_script(fio_data_file,title,gnuplot_output_filename,mode,disk_perf): |
| 41 | f=open("mygraph",'w') |
| 42 | if len(fio_data_file) > 1: |
| 43 | f.write("call \'graph3D.gpm\' \'%s' \'%s\' \'\' \'%s\' \'%s\'\n" % (title,gnuplot_output_filename,gnuplot_output_filename,mode)) |
| 44 | |
| 45 | pos=0 |
| 46 | # Let's create a temporary file for each selected fio file |
| 47 | for file in fio_data_file: |
| 48 | tmp_filename = "gnuplot_temp_file.%d" % pos |
| 49 | png_file=file.replace('.log','') |
| 50 | raw_filename = "%s-2Draw" % (png_file) |
| 51 | smooth_filename = "%s-2Dsmooth" % (png_file) |
| 52 | trend_filename = "%s-2Dtrend" % (png_file) |
| 53 | avg = average(disk_perf[pos]) |
| 54 | f.write("call \'graph2D.gpm\' \'%s' \'%s\' \'\' \'%s\' \'%s\' \'%s\' \'%s\' \'%f\'\n" % (title,tmp_filename,raw_filename,mode,smooth_filename,trend_filename,avg)) |
| 55 | pos = pos +1 |
| 56 | |
| 57 | f.close() |
| 58 | |
| 59 | def generate_gnuplot_math_script(title,gnuplot_output_filename,mode,average): |
| 60 | f=open("mymath",'a') |
| 61 | f.write("call \'math.gpm\' \'%s' \'%s\' \'\' \'%s\' \'%s\' %s\n" % (title,gnuplot_output_filename,gnuplot_output_filename,mode,average)) |
| 62 | f.close() |
| 63 | |
| 64 | def compute_aggregated_file(fio_data_file, gnuplot_output_filename): |
| 65 | temp_files=[] |
| 66 | pos=0 |
| 67 | # Let's create a temporary file for each selected fio file |
| 68 | for file in fio_data_file: |
| 69 | tmp_filename = "gnuplot_temp_file.%d" % pos |
| 70 | temp_files.append(open(tmp_filename,'r')) |
| 71 | pos = pos +1 |
| 72 | |
| 73 | f = open(gnuplot_output_filename, "w") |
| 74 | index=0 |
| 75 | # Let's add some information |
| 76 | for tempfile in temp_files: |
| 77 | f.write("# Disk%d was coming from %s\n" % (index,fio_data_file[index])) |
| 78 | f.write(tempfile.read()) |
| 79 | f.write("\n") |
| 80 | tempfile.close() |
| 81 | index = index + 1 |
| 82 | f.close() |
| 83 | |
| 84 | def average(s): return sum(s) * 1.0 / len(s) |
| 85 | |
| 86 | def compute_temp_file(fio_data_file,disk_perf): |
| 87 | files=[] |
| 88 | temp_outfile=[] |
| 89 | blk_size=0 |
| 90 | for file in fio_data_file: |
| 91 | files.append(open(file)) |
| 92 | pos = len(files) - 1 |
| 93 | tmp_filename = "gnuplot_temp_file.%d" % pos |
| 94 | gnuplot_file=open(tmp_filename,'w') |
| 95 | temp_outfile.append(gnuplot_file) |
| 96 | gnuplot_file.write("#Temporary file based on file %s\n" % file) |
| 97 | disk_perf.append([]) |
| 98 | |
| 99 | shall_break = False |
| 100 | while True: |
| 101 | current_line=[] |
| 102 | nb_empty_files=0 |
| 103 | nb_files=len(files) |
| 104 | for file in files: |
| 105 | s=file.readline().replace(',',' ').split() |
| 106 | if not s: |
| 107 | nb_empty_files+=1 |
| 108 | s="-1, 0, 0, 0'".replace(',',' ').split() |
| 109 | |
| 110 | if (nb_empty_files == nb_files): |
| 111 | shall_break=True |
| 112 | break; |
| 113 | |
| 114 | current_line.append(s); |
| 115 | |
| 116 | if shall_break == True: |
| 117 | break |
| 118 | |
| 119 | last_time = -1 |
| 120 | index=0 |
| 121 | perfs=[] |
| 122 | for line in current_line: |
| 123 | time, perf, x, block_size = line |
| 124 | if (blk_size == 0): |
| 125 | blk_size=int(block_size) |
| 126 | |
| 127 | # We ignore the first 500msec as it doesn't seems to be part of the real benchmark |
| 128 | # Time < 500 usually reports BW=0 breaking the min computing |
| 129 | if (((int(time)) > 500) or (int(time)==-1)): |
| 130 | disk_perf[index].append(int(perf)) |
| 131 | perfs.append("%s %s"% (time, perf)) |
| 132 | index = index + 1 |
| 133 | |
| 134 | # If we reach this point, it means that all the traces are coherent |
| 135 | for p in enumerate(perfs): |
| 136 | perf_time,perf = p[1].split() |
| 137 | if (perf_time != "-1"): |
| 138 | temp_outfile[p[0]].write("%s %.2f %s\n" % (p[0], float(float(perf_time)/1000), perf)) |
| 139 | |
| 140 | |
| 141 | for file in files: |
| 142 | file.close() |
| 143 | for file in temp_outfile: |
| 144 | file.close() |
| 145 | return blk_size |
| 146 | |
| 147 | def compute_math(fio_data_file, title,gnuplot_output_filename,mode,disk_perf): |
| 148 | global_min=[] |
| 149 | global_max=[] |
| 150 | average_file=open(gnuplot_output_filename+'.average', 'w') |
| 151 | min_file=open(gnuplot_output_filename+'.min', 'w') |
| 152 | max_file=open(gnuplot_output_filename+'.max', 'w') |
| 153 | stddev_file=open(gnuplot_output_filename+'.stddev', 'w') |
| 154 | global_file=open(gnuplot_output_filename+'.global','w') |
| 155 | |
| 156 | min_file.write('DiskName %s\n' % mode) |
| 157 | max_file.write('DiskName %s\n'% mode) |
| 158 | average_file.write('DiskName %s\n'% mode) |
| 159 | stddev_file.write('DiskName %s\n'% mode ) |
| 160 | for disk in xrange(len(fio_data_file)): |
| 161 | # print disk_perf[disk] |
| 162 | min_file.write("# Disk%d was coming from %s\n" % (disk,fio_data_file[disk])) |
| 163 | max_file.write("# Disk%d was coming from %s\n" % (disk,fio_data_file[disk])) |
| 164 | average_file.write("# Disk%d was coming from %s\n" % (disk,fio_data_file[disk])) |
| 165 | stddev_file.write("# Disk%d was coming from %s\n" % (disk,fio_data_file[disk])) |
| 166 | avg = average(disk_perf[disk]) |
| 167 | variance = map(lambda x: (x - avg)**2, disk_perf[disk]) |
| 168 | standard_deviation = math.sqrt(average(variance)) |
| 169 | # print "Disk%d [ min=%.2f max=%.2f avg=%.2f stddev=%.2f \n" % (disk,min(disk_perf[disk]),max(disk_perf[disk]),avg, standard_deviation) |
| 170 | average_file.write('%d %d\n' % (disk, avg)) |
| 171 | stddev_file.write('%d %d\n' % (disk, standard_deviation)) |
| 172 | local_min=min(disk_perf[disk]) |
| 173 | local_max=max(disk_perf[disk]) |
| 174 | min_file.write('%d %d\n' % (disk, local_min)) |
| 175 | max_file.write('%d %d\n' % (disk, local_max)) |
| 176 | global_min.append(int(local_min)) |
| 177 | global_max.append(int(local_max)) |
| 178 | |
| 179 | global_disk_perf = sum(disk_perf, []) |
| 180 | avg = average(global_disk_perf) |
| 181 | variance = map(lambda x: (x - avg)**2, global_disk_perf) |
| 182 | standard_deviation = math.sqrt(average(variance)) |
| 183 | |
| 184 | global_file.write('min=%.2f\n' % min(global_disk_perf)) |
| 185 | global_file.write('max=%.2f\n' % max(global_disk_perf)) |
| 186 | global_file.write('avg=%.2f\n' % avg) |
| 187 | global_file.write('stddev=%.2f\n' % standard_deviation) |
| 188 | global_file.write('values_count=%d\n' % len(global_disk_perf)) |
| 189 | global_file.write('disks_count=%d\n' % len(fio_data_file)) |
| 190 | #print "Global [ min=%.2f max=%.2f avg=%.2f stddev=%.2f \n" % (min(global_disk_perf),max(global_disk_perf),avg, standard_deviation) |
| 191 | |
| 192 | average_file.close() |
| 193 | min_file.close() |
| 194 | max_file.close() |
| 195 | stddev_file.close() |
| 196 | global_file.close() |
| 197 | try: |
| 198 | os.remove('mymath') |
| 199 | except: |
| 200 | True |
| 201 | |
| 202 | generate_gnuplot_math_script("Average values of "+title,gnuplot_output_filename+'.average',mode,int(avg)) |
| 203 | generate_gnuplot_math_script("Min values of "+title,gnuplot_output_filename+'.min',mode,average(global_min)) |
| 204 | generate_gnuplot_math_script("Max values of "+title,gnuplot_output_filename+'.max',mode,average(global_max)) |
| 205 | generate_gnuplot_math_script("Standard Deviation of "+title,gnuplot_output_filename+'.stddev',mode,int(standard_deviation)) |
| 206 | |
| 207 | def parse_global_files(fio_data_file, global_search): |
| 208 | max_result=0 |
| 209 | max_file='' |
| 210 | for file in fio_data_file: |
| 211 | f=open(file) |
| 212 | disk_count=0 |
| 213 | search_value=-1 |
| 214 | |
| 215 | # Let's read the complete file |
| 216 | while True: |
| 217 | try: |
| 218 | # We do split the name from the value |
| 219 | name,value=f.readline().split("=") |
| 220 | except: |
| 221 | f.close() |
| 222 | break |
| 223 | # If we ended the file |
| 224 | if not name: |
| 225 | # Let's process what we have |
| 226 | f.close() |
| 227 | break |
| 228 | else: |
| 229 | # disks_count is not global_search item |
| 230 | # As we need it for some computation, let's save it |
| 231 | if name=="disks_count": |
| 232 | disks_count=int(value) |
| 233 | |
| 234 | # Let's catch the searched item |
| 235 | if global_search in name: |
| 236 | search_value=float(value) |
| 237 | |
| 238 | # Let's process the avg value by estimated the global bandwidth per file |
| 239 | # We keep the biggest in memory for reporting |
| 240 | if global_search == "avg": |
| 241 | if (disks_count > 0) and (search_value != -1): |
| 242 | result=disks_count*search_value |
| 243 | if (result > max_result): |
| 244 | max_result=result |
| 245 | max_file=file |
| 246 | # Let's print the avg output |
| 247 | if global_search == "avg": |
| 248 | print "Biggest aggregated value of %s was %2.f in file %s\n" % (global_search, max_result, max_file) |
| 249 | else: |
| 250 | print "Global search %s is not yet implemented\n" % global_search |
| 251 | |
| 252 | def render_gnuplot(): |
| 253 | print "Running gnuplot Rendering\n" |
| 254 | try: |
| 255 | os.system("gnuplot mymath") |
| 256 | os.system("gnuplot mygraph") |
| 257 | except: |
| 258 | print "Could not run gnuplot on mymath or mygraph !\n" |
| 259 | sys.exit(1); |
| 260 | |
| 261 | def print_help(): |
| 262 | print 'fio2gnuplot.py -ghbio -t <title> -o <outputfile> -p <pattern>' |
| 263 | print |
| 264 | print '-h --help : Print this help' |
| 265 | print '-p <pattern> or --pattern <pattern> : A pattern in regexp to select fio input files' |
| 266 | print '-b or --bandwidth : A predefined pattern for selecting *_bw.log files' |
| 267 | print '-i or --iops : A predefined pattern for selecting *_iops.log files' |
| 268 | print '-g or --gnuplot : Render gnuplot traces before exiting' |
| 269 | print '-o or --outputfile <file> : The basename for gnuplot traces' |
| 270 | print ' - Basename is set with the pattern if defined' |
| 271 | print '-t or --title <title> : The title of the gnuplot traces' |
| 272 | print ' - Title is set with the block size detected in fio traces' |
| 273 | print '-G or --Global <type> : Search for <type> in .global files match by a pattern' |
| 274 | print ' - Available types are : min, max, avg, stddev' |
| 275 | print ' - The .global extension is added automatically to the pattern' |
| 276 | |
| 277 | def main(argv): |
| 278 | mode='unknown' |
| 279 | pattern='' |
| 280 | pattern_set_by_user=False |
| 281 | title='No title' |
| 282 | gnuplot_output_filename='result' |
| 283 | disk_perf=[] |
| 284 | run_gnuplot=False |
| 285 | parse_global=False |
| 286 | global_search='' |
| 287 | |
| 288 | try: |
| 289 | opts, args = getopt.getopt(argv[1:],"ghbio:t:p:G:") |
| 290 | except getopt.GetoptError: |
| 291 | print_help() |
| 292 | sys.exit(2) |
| 293 | |
| 294 | for opt, arg in opts: |
| 295 | if opt in ("-b", "--bandwidth"): |
| 296 | pattern='*_bw.log' |
| 297 | elif opt in ("-i", "--iops"): |
| 298 | pattern='*_iops.log' |
| 299 | elif opt in ("-p", "--pattern"): |
| 300 | pattern_set_by_user=True |
| 301 | pattern=arg |
| 302 | pattern=pattern.replace('\\','') |
| 303 | elif opt in ("-o", "--outputfile"): |
| 304 | gnuplot_output_filename=arg |
| 305 | elif opt in ("-t", "--title"): |
| 306 | title=arg |
| 307 | elif opt in ("-g", "--gnuplot"): |
| 308 | run_gnuplot=True |
| 309 | elif opt in ("-G", "--Global"): |
| 310 | parse_global=True |
| 311 | global_search=arg |
| 312 | elif opt in ("-h", "--help"): |
| 313 | print_help() |
| 314 | sys.exit(1) |
| 315 | |
| 316 | # Adding .global extension to the file |
| 317 | if parse_global==True: |
| 318 | if not gnuplot_output_filename.endswith('.global'): |
| 319 | pattern = pattern+'.global' |
| 320 | |
| 321 | fio_data_file=find_file('.',pattern) |
| 322 | if len(fio_data_file) == 0: |
| 323 | print "No log file found with pattern %s!" % pattern |
| 324 | sys.exit(1) |
| 325 | |
| 326 | fio_data_file=sorted(fio_data_file, key=str.lower) |
| 327 | for file in fio_data_file: |
| 328 | print 'Selected %s' % file |
| 329 | if "_bw.log" in file : |
| 330 | mode="Bandwidth (KB/sec)" |
| 331 | if "_iops.log" in file : |
| 332 | mode="IO per Seconds (IO/sec)" |
| 333 | if (title == 'No title') and (mode != 'unknown'): |
| 334 | if "Bandwidth" in mode: |
| 335 | title='Bandwidth benchmark with %d fio results' % len(fio_data_file) |
| 336 | if "IO" in mode: |
| 337 | title='IO benchmark with %d fio results' % len(fio_data_file) |
| 338 | |
| 339 | #We need to adjust the output filename regarding the pattern required by the user |
| 340 | if (pattern_set_by_user == True): |
| 341 | gnuplot_output_filename=pattern |
| 342 | # As we do have some regexp in the pattern, let's make this simpliest |
| 343 | # We do remove the simpliest parts of the expression to get a clear file name |
| 344 | gnuplot_output_filename=gnuplot_output_filename.replace('-*-','-') |
| 345 | gnuplot_output_filename=gnuplot_output_filename.replace('*','-') |
| 346 | gnuplot_output_filename=gnuplot_output_filename.replace('--','-') |
| 347 | gnuplot_output_filename=gnuplot_output_filename.replace('.log','') |
| 348 | # Insure that we don't have any starting or trailing dash to the filename |
| 349 | gnuplot_output_filename = gnuplot_output_filename[:-1] if gnuplot_output_filename.endswith('-') else gnuplot_output_filename |
| 350 | gnuplot_output_filename = gnuplot_output_filename[1:] if gnuplot_output_filename.startswith('-') else gnuplot_output_filename |
| 351 | |
| 352 | if parse_global==True: |
| 353 | parse_global_files(fio_data_file, global_search) |
| 354 | else: |
| 355 | blk_size=compute_temp_file(fio_data_file,disk_perf) |
| 356 | title="%s @ Blocksize = %dK" % (title,blk_size/1024) |
| 357 | compute_aggregated_file(fio_data_file, gnuplot_output_filename) |
| 358 | compute_math(fio_data_file,title,gnuplot_output_filename,mode,disk_perf) |
| 359 | generate_gnuplot_script(fio_data_file,title,gnuplot_output_filename,mode,disk_perf) |
| 360 | |
| 361 | if (run_gnuplot==True): |
| 362 | render_gnuplot() |
| 363 | |
| 364 | # Cleaning temporary files |
| 365 | try: |
| 366 | os.remove('gnuplot_temp_file.*') |
| 367 | except: |
| 368 | True |
| 369 | |
| 370 | #Main |
| 371 | if __name__ == "__main__": |
| 372 | sys.exit(main(sys.argv)) |