Gheek.net

September 12, 2012

iotop2csv – One way to graph the batch output of iotop

Filed under: linux, perl — lancevermilion @ 5:38 pm

I had a need to know what was writing disk, how much it was writing, and how often it was writing. iotop does a pretty good job of giving you a top like view of disk I/O activity. I decided I wanted to graph this to hopefully make it easier for me to see how heavy writing to the disk might correlate to sluggish CLI response.

The output is in CSV so you can drop it in Excel (if you so wish) and create a simple PivotChart with it. You can do so by selecting all data from the csv (after it is imported/converted to rows/columns using data import or text to columns) then choose insert pivotchart.

Choose fields timestamp, tid, read, write and cmd.
Timestamp goes in AxisFields Categories.
Values, cmd, and tid go in Legend Fields
Sum of read and Sum of write goes in Values.

I never quite got to the smoking gun (just yet) but I was able to see what applications were busy little apps writing to the disk.

I do no warranty this code in any way. Use it as you wish. If you find it useful then please write back saying so. If you find it incorrect or improve/fix it please write back on that as well.

Here is the code

#/usr/bin/perl -w
use strict;
# Enabling strict with strict refs breaks the summary output (when ctrl+c is pressed)
# because of the error Can't use string ("") as a HASH ref while "strict refs" in use.
no strict 'refs';

#Initialize an empty hashref
my $hash_ref = {};

#Catch CTRL+C and then call printsummary subroutine
$SIG{'INT'} = \&printsummary;

#Pipe the output from iotop to a file handle
open fh, "iotop -bt|" or die $!;

# Build the initial Total Value key/values
$hash_ref->{'tsample'} = 0;
$hash_ref->{'tread'} = 0;
$hash_ref->{'twrite'} = 0;

# Divider of columns
my $div = ",";

#Print the header
print "#" x 50 . "\n";
print "Output from iotop -bt\n";
print "#" x 50 . "\n";
print "timestamp" . $div . "tid" . $div . "user" . $div . "read" . $div . "total read each tid" . $div . "write" . $div . "total write each tid" . $div . "cmd" . $div . "sample number" . $div . "Percent of Total Read" . $div . "Percent of Total Write" . "\n";

#While data from iotop process it. Since we get lots of 0.00 read and write values
#don't waste time storing/printing values that are useless. Only store lines that
#have a value to read/write.
while(<fh>) {
  my ($ts,$tid,$prio,$user,$read,undef,$write,undef,undef,undef,undef,undef,$cmd) = split(/\s+/, "$_");
  chomp($cmd);

  #Make sure we fill each tid with dummy values if the tid is a real valid number
  if ( !$hash_ref->{$tid} && ( $tid =~ m/^-?\d+$/ || $tid =~ m/^-?\d+[\/|\.]\d+$/ ) )
  {
    $hash_ref->{$tid}->{'user'} = 'null';
    $hash_ref->{$tid}->{'read'} = 0;
    $hash_ref->{$tid}->{'write'} = 0;
    $hash_ref->{$tid}->{'cmd'} = 'null';
    $hash_ref->{$tid}->{'samples'} = 0;
  }

  #Continue if we have already created a key for the tid and the tid is a real valid number
  if ( $hash_ref->{$tid} && ( $tid =~ m/^-?\d+$/ || $tid =~ m/^-?\d+[\/|\.]\d+$/ ) )
  {
    $hash_ref->{$tid}->{'user'} = $user;
    $hash_ref->{$tid}->{'read'} += $read if $read =~ m/^-?\d+$/ || $read =~ m/^-?\d+[\/|\.]\d+$/;
    $hash_ref->{$tid}->{'write'} += $write if $write =~ m/^-?\d+$/ || $write =~ m/^-?\d+[\/|\.]\d+$/;
    $hash_ref->{$tid}->{'cmd'} = $cmd;
    $hash_ref->{$tid}->{'samples'}++;

    # create totals key/value pair
    $hash_ref->{'tsample'}++;
    $hash_ref->{'tread'} += $read if $read =~ m/^-?\d+$/ || $read =~ m/^-?\d+[\/|\.]\d+$/;
    $hash_ref->{'twrite'} += $write if $write =~ m/^-?\d+$/ || $write =~ m/^-?\d+[\/|\.]\d+$/;

    # Print values of the read/write as they happen
    if ( ( $read =~ m/^-?\d+$/ || $read =~ m/^-?\d+[\/|\.]\d+$/ || $write =~ m/^-?\d+$/ || $write =~ m/^-?\d+[\/|\.]\d+$/ ) && ( $read > 0 || $write > 0 ) )
    {
      my $ptr = 0;
      my $ptw = 0;
      $ptr = ( $hash_ref->{$tid}->{'read'} / $hash_ref->{'tread'} ) * 100 if ( $read > 0 ) && ( $hash_ref->{'tread'} > 0 );
      $ptw = ( $hash_ref->{$tid}->{'write'} / $hash_ref->{'twrite'} ) * 100 if ( $write > 0 ) && ( $hash_ref->{'twrite'} > 0 );

      print $ts . "$div";
      print $tid . "$div";
      print $hash_ref->{$tid}->{'user'} . "$div";
      print $read . "$div";
      print "$hash_ref->{$tid}->{'read'}" . "$div";
      print $write . "$div";
      print "$hash_ref->{$tid}->{'write'}" . "$div";
      print $hash_ref->{$tid}->{'cmd'} . "$div";
      print $hash_ref->{$tid}->{'samples'} . "$div";
      print sprintf("%.2f", $ptr) . "$div";
      print sprintf("%.2f", $ptw) . "\n";
    }
  }
}

# Close the filehandle when we are done.
close(fh);

#Sub routine to print out a summary of disk I/O.
sub printsummary {
print "#" x 50 . "\n";
print "Caught Ctrl+C. Printing Disk I/O summary.\n";
print "#" x 50 . "\n";
print "Total Bytes Read:  $hash_ref->{'tread'}\n";
print "Total Bytes Write: $hash_ref->{'twrite'}\n";
print "#" x 50 . "\n";

for my $id ( keys %$hash_ref )
  {
  if ( $hash_ref->{$id}->{'read'} > 0 || $hash_ref->{$id}->{'write'} > 0 )
    {
      my $psptr = 0;
      my $psptw = 0;
      $psptr = ( $hash_ref->{$id}->{'read'} / $hash_ref->{'tread'} ) * 100 if ( $hash_ref->{'tread'} > 0 );
      $psptw = ( $hash_ref->{$id}->{'write'} / $hash_ref->{'twrite'} ) * 100 if ( $hash_ref->{'twrite'} > 0 );

      print "$id" . " ";
      print $hash_ref->{$id}->{'user'} . " ";
      print $hash_ref->{$id}->{'read'} . " ";
      print $hash_ref->{$id}->{'write'} . " ";
      print $hash_ref->{$id}->{'cmd'} . " ";
      print $hash_ref->{$id}->{'samples'} . " ";
      print sprintf("%.2f", $psptr) . " ";
      print sprintf("%.2f", $psptw);
      print "\n"
    }
  }
  print "#" x 50 . "\n";
}

Here is the csv output

sudo perl iotop2csv.pl 
##################################################
Output from iotop -bt
##################################################
timestamp,tid,user,read,total read each tid,write,total write each tid,cmd,sample number,Percent of Total Read,Percent of Total Write
16:27:00,6003,mysql,0.00,0,318.04,318.04,mysqld,2,0.00,100.00
16:27:00,4959,root,0.00,0,3.74,3.74,nailslogd,2,0.00,1.16
16:27:00,5992,mysql,0.00,0,3.74,3.74,mysqld,2,0.00,1.15
16:27:00,5994,mysql,0.00,0,314.30,314.3,mysqld,2,0.00,49.12
16:27:01,511,root,0.00,0,101.83,101.83,[kjournald],3,0.00,13.73
16:27:01,6557,mysql,0.00,0,3.77,3.77,mysqld,3,0.00,0.51
16:27:02,6530,mysql,0.00,0,7.55,7.55,mysqld,4,0.00,1.00
16:27:03,6009,mysql,0.00,0,7.54,7.54,mysqld,5,0.00,0.99
16:27:03,25220,mysql,3.77,3.77,11.31,11.31,mysqld,5,100.00,1.47
16:27:04,6534,mysql,3.77,3.77,97.94,97.94,mysqld,6,50.00,11.26
16:27:04,1864,root,0.00,0,199.65,199.65,[kjournald],6,0.00,18.67

Here is an example graph via Excel

Advertisements

1 Comment »

  1. Reblogged this on Gigable – Tech Blog.

    Comment by Bosstiger — September 12, 2012 @ 11:42 pm


RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at WordPress.com.

%d bloggers like this: