fio2gnuplot and iolog fixes

Dear Jens,

Please find below my latest works for fio.
I do have some fio2gnuplot fixes and one is very important as some data were mixed leading to weird/stupid plotting artefacts like loops....

Aside this, I've been fixing the empty field when logging IOs.

The following changes since commit 428b4f6ba681dbb40c8e2213d0c6ae8f8049dcd5:

fio2gnuplot: Adding filename as title for 2D graphs (2013-07-31
12:32:40 +0200)

are available in the git repository at:

[email protected]:enovance/fio.git erwan/fixes

for you to fetch changes up to aab5954b98a275ca1cf29d5339cdc319d9f4cfda:

iolog: Logging blocksize in IOPS traces (2013-08-02 16:39:40 +0200)

----------------------------------------------------------------
Erwan Velu (8):
fio2gnuplot: Fixing typo when inserting fake data
fio2gnuplot: Print line if int() conversion fails
fio2gnuplot: Adding time selection options
fio2gnplot: Renaming file variable to myfile
fio2gnuplot: Fixing plotting issues on shorten files
fio2gnuplot: Plotting average value for compare graphs
fio2gnuplot: Display average value in label
iolog: Logging blocksize in IOPS traces

 io_u.c                    |  2 +-
 iolog.h                   |  3 ++-
 stat.c                    |  4 ++--
 tools/plot/fio2gnuplot.py | 79 
+++++++++++++++++++++++++++++++++++++++++++++++++++----------------------------
 tools/plot/graph2D.gpm    |  6 +++---
 tools/plot/math.gpm       |  2 +-
 6 files changed, 60 insertions(+), 36 deletions(-)

index 6537c90..a35aafd 100644
@@ -1412,7 +1412,7 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u,
     if (!td->o.disable_bw)
         add_bw_sample(td, idx, bytes, &icd->time);
 
-    add_iops_sample(td, idx, &icd->time);
+    add_iops_sample(td, idx, bytes, &icd->time);
 }
 
 static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)

index 7099097..8fedc19 100644
@@ -118,7 +118,8 @@ extern void add_slat_sample(struct thread_data *, enum fio_ddir, unsigned long,
                 unsigned int);
 extern void add_bw_sample(struct thread_data *, enum fio_ddir, 

unsigned int,

                 struct timeval *);

-extern void add_iops_sample(struct thread_data *, enum fio_ddir, struct timeval *); +extern void add_iops_sample(struct thread_data *, enum fio_ddir, unsigned int,
+ struct timeval *);
extern void init_disk_util(struct thread_data *);
extern void update_rusage_stat(struct thread_data *);
extern void setup_log(struct io_log **, unsigned long, int);

index 442caa0..10d9efe 100644
@@ -1691,7 +1691,7 @@ void add_bw_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs,
     fio_gettime(&td->bw_sample_time, NULL);
 }
 
-void add_iops_sample(struct thread_data *td, enum fio_ddir ddir,
+void add_iops_sample(struct thread_data *td, enum fio_ddir ddir, unsigned int bs,
              struct timeval *t)
 {
     struct thread_stat *ts = &td->ts;

@@ -1718,7 +1718,7 @@ void add_iops_sample(struct thread_data *td, enum fio_ddir ddir,
         add_stat_sample(&ts->iops_stat[ddir], iops);
 
         if (td->iops_log)
-            add_log_sample(td, td->iops_log, iops, ddir, 0);
+            add_log_sample(td, td->iops_log, iops, ddir, bs);
 
         td->stat_io_blocks[ddir] = td->this_io_blocks[ddir];
     }

index 800c334..64b41c1 100755
@@ -56,8 +56,8 @@ set ytics axis out auto
 set key top left reverse
 set xlabel "Time (Seconds)"
 set ylabel '%s'
-set xrange [0:]
 set yrange [0:]
+set style line 1 lt 1 lw 3 pt 3 linecolor rgb "green"
 '''% (title,mode))
         compare.close()
         #Copying the common file for all kind of graph (raw/smooth/trend)

@@ -76,6 +76,13 @@ set yrange [0:]
 compare_trend=open(gnuplot_output_dir+compare_trend_filename+".gnuplot",'a')
         compare_trend.write("set output '%s.png'\n" % 

compare_trend_filename)

+        # Let's plot the average value for all the traces
+        global_disk_perf = sum(disk_perf, [])
+        global_avg  = average(global_disk_perf)
+        compare_raw.write("plot %s w l ls 1 ti 'Global average value (%.2f)'" % (global_avg,global_avg));
+        compare_smooth.write("plot %s w l ls 1 ti 'Global average value (%.2f)'" % (global_avg,global_avg));
+        compare_trend.write("plot %s w l ls 1 ti 'Global average value (%.2f)'" % (global_avg,global_avg));

+
pos=0
# Let's create a temporary file for each selected fio file for file in fio_data_file:
@@ -84,14 +91,9 @@ set yrange [0:]
# Plotting comparing graphs doesn't have a meaning unless if
there is at least 2 traces
if len(fio_data_file) > 1:
# Adding the plot instruction for each kind of comparing
graphs
- if pos ==0 :

-                compare_raw.write("plot '%s' using 2:3 with linespoints title '%s'" % (tmp_filename,fio_data_file[pos]))
-                compare_smooth.write("plot '%s' using 2:3 smooth csplines title '%s'" % (tmp_filename,fio_data_file[pos]))
-                compare_trend.write("plot '%s' using 2:3 smooth bezier title '%s'" % (tmp_filename,fio_data_file[pos]))
-            else:
-                compare_raw.write(",\\\n'%s' using 2:3 with linespoints title '%s'" % (tmp_filename,fio_data_file[pos]))
-                compare_smooth.write(",\\\n'%s' using 2:3 smooth csplines title '%s'" % (tmp_filename,fio_data_file[pos]))
-                compare_trend.write(",\\\n'%s' using 2:3 smooth bezier title '%s'" % (tmp_filename,fio_data_file[pos]))
+            compare_raw.write(",\\\n'%s' using 2:3 with linespoints title '%s'" % (tmp_filename,fio_data_file[pos]))
+            compare_smooth.write(",\\\n'%s' using 2:3 smooth csplines title '%s'" % (tmp_filename,fio_data_file[pos]))
+            compare_trend.write(",\\\n'%s' using 2:3 smooth bezier title '%s'" % (tmp_filename,fio_data_file[pos]))

         png_file=file.replace('.log','')
                 raw_filename = "%s-2Draw" % (png_file)

@@ -139,8 +141,11 @@ def compute_aggregated_file(fio_data_file, gnuplot_output_filename, gnuplot_outp

def average(s): return sum(s) * 1.0 / len(s)

-def compute_temp_file(fio_data_file,disk_perf,gnuplot_output_dir):
- print "Processing data file 1/2"
+def compute_temp_file(fio_data_file,disk_perf,gnuplot_output_dir, min_time, max_time):
+ end_time=max_time
+ if end_time == -1:
+ end_time="infinite"
+ print "Processing data file 1/2 with %s<time<%s" % (min_time,end_time)

     files=[]
     temp_outfile=[]
     blk_size=0
@@ -158,11 +163,11 @@ def compute_temp_file(fio_data_file,disk_perf,gnuplot_output_dir):
         current_line=[]
         nb_empty_files=0
         nb_files=len(files)

- for file in files:
- s=file.readline().replace(',',' ').split()
+ for myfile in files:
+ s=myfile.readline().replace(',',' ').split()
if not s:
nb_empty_files+=1
- s="-1, 0, 0, 0'".replace(',',' ').split()
+ s="-1, 0, 0, 0".replace(',',' ').split()

             if (nb_empty_files == nb_files):
                 shall_break=True
@@ -174,25 +179,35 @@ def compute_temp_file(fio_data_file,disk_perf,gnuplot_output_dir):
             break

         last_time = -1
-        index=0
+        index=-1
         perfs=[]

- for line in current_line:
- time, perf, x, block_size = line
+ for line in enumerate(current_line):
+ # Index will be used to remember what file was featuring what value
+ index=index+1
+
+ time, perf, x, block_size = line[1]

             if (blk_size == 0):

- blk_size=int(block_size)
+ try:
+ blk_size=int(block_size)
+ except:

+                    print "Error while reading the following line :"
+                    print line
+                    sys.exit(1);

# We ignore the first 500msec as it doesn't seems to be
part of the real benchmark
# Time < 500 usually reports BW=0 breaking the min computing
- if (((int(time)) > 500) or (int(time)==-1)):

-                disk_perf[index].append(int(perf))
-                perfs.append("%s %s"% (time, perf))
-                index = index + 1

+ if (min_time == 0):
+ min_time==0.5
+
+ # Then we estimate if the data we got is part of the time range we want to plot
+ if ((float(time)>(float(min_time)*1000)) and ((int(time) < (int(max_time)*1000)) or max_time==-1)):
+ disk_perf[index].append(int(perf))
+ perfs.append("%d %s %s"% (index, time, perf))

# If we reach this point, it means that all the traces are
coherent
for p in enumerate(perfs):
- perf_time,perf = p[1].split()
- if (perf_time != "-1"):
- temp_outfile[p[0]].write("%s %.2f %s\n" % (p[0], float(float(perf_time)/1000), perf))
+ index, perf_time,perf = p[1].split()
+ temp_outfile[int(index)].write("%s %.2f %s\n" % (index, float(float(perf_time)/1000), perf))

for file in files:
@@ -328,7 +343,7 @@ def render_gnuplot(fio_data_file, gnuplot_output_dir):

         sys.exit(1);

def print_help():
- print 'fio2gnuplot.py -ghbiod -t <title> -o <outputfile> -p <pattern> -G <type>'
+ print 'fio2gnuplot.py -ghbiod -t <title> -o <outputfile> -p <pattern> -G <type> -m <time> -M <time>'

     print
     print '-h --help                           : Print this help'
     print '-p <pattern> or --pattern <pattern> : A pattern in regexp 
to select fio input files'
@@ -343,6 +358,8 @@ def print_help():
     print '-G        or --Global <type>     : Search for <type> in 

.global files match by a pattern'
print ' - Available types
are : min, max, avg, stddev'
print ' - The .global
extension is added automatically to the pattern'

+    print '-m           or --min_time <time>   : Only consider data starting from <time> seconds (default is 0)'
+    print '-M           or --max_time <time>   : Only consider data ending before <time> seconds (default is -1 aka nolimit)'

def main(argv):
mode='unknown'
@@ -356,6 +373,8 @@ def main(argv):
run_gnuplot=False
parse_global=False
global_search=''
+ min_time=0
+ max_time=-1

if not os.path.isfile(gpm_dir+'math.gpm'):
gpm_dir="/usr/local/share/fio/"
@@ -364,7 +383,7 @@ def main(argv):

             sys.exit(3)

     try:
-        opts, args = getopt.getopt(argv[1:],"ghbio:d:t:p:G:")
+        opts, args = getopt.getopt(argv[1:],"ghbio:d:t:p:G:m:M:")
     except getopt.GetoptError:
      print_help()
          sys.exit(2)
@@ -388,6 +407,10 @@ def main(argv):
         os.makedirs(gnuplot_output_dir)
       elif opt in ("-t", "--title"):
          title=arg

+ elif opt in ("-m", "--min_time"):
+ min_time=arg
+ elif opt in ("-M", "--max_time"):
+ max_time=arg

       elif opt in ("-g", "--gnuplot"):
      run_gnuplot=True
       elif opt in ("-G", "--Global"):
@@ -439,7 +462,7 @@ def main(argv):
     if parse_global==True:
     parse_global_files(fio_data_file, global_search)
     else:
-  blk_size=compute_temp_file(fio_data_file,disk_perf,gnuplot_output_dir)
+ 
 blk_size=compute_temp_file(fio_data_file,disk_perf,gnuplot_output_dir,min_time,max_time)
         title="%s @ Blocksize = %dK" % (title,blk_size/1024)
         compute_aggregated_file(fio_data_file, 

gnuplot_output_filename, gnuplot_output_dir)
compute_math(fio_data_file,title,gnuplot_output_filename,gnuplot_output_dir,mode,disk_perf,gpm_dir)

index efc7ac2..5cd6ff3 100644
@@ -22,13 +22,13 @@ set yrange [0:]
 set style line 100 lt 7 lw 0.5
 set style line 1 lt 1 lw 3 pt 3 linecolor rgb "green"
 
-plot '$1' using 2:3 with linespoints title '$2', $7 w l ls 1 ti 'Global average value'
+plot '$1' using 2:3 with linespoints title '$2', $7 w l ls 1 ti 'Global average value ($7)'
 
 set output '$5.png'
-plot '$1' using 2:3 smooth csplines title '$2', $7 w l ls 1 ti 'Global average value'
+plot '$1' using 2:3 smooth csplines title '$2', $7 w l ls 1 ti 'Global average value ($7)'
 
 set output '$6.png'
-plot '$1' using 2:3 smooth bezier title '$2', $7 w l ls 1 ti 'Global average value'
+plot '$1' using 2:3 smooth bezier title '$2', $7 w l ls 1 ti 'Global average value ($7)'
 
 #pause -1
 #The End

index 81e03da..a01f5a0 100644
@@ -22,4 +22,4 @@ set xtics axis out
 set xtic rotate by 45 scale 0 font ",8" autojustify
 set xtics offset 0,-1 border -5,1,5
 set style line 1 lt 1 lw 3 pt 3 linecolor rgb "green"
-plot '$1' using 2:xtic(1) ti col, $5 w l ls 1 ti 'Global average value'
+plot '$1' using 2:xtic(1) ti col, $5 w l ls 1 ti 'Global average value ($5)'