#!/usr/bin/perl
#display helpfile
if ($ARGV[0] eq "-help") {
    print "\nThis script calculates the usage statistics for linux cluster from the Rtop\n"; 
    print "summary files. The results are summarized in HTML format. The main page is\n"; 
    print "data.html\n\n";
    print "The script must be called from the directory containing the Rtop files to\n";
    print "be summarized\n\n";
    print "usage is vnAccount [-p <percent>] [-r <dirname>] [-help]\n\n";
    print "-p: sets the maxmimum value of the horizontal scale for the bar graphs.\n";
    print "-p must be followed by a number between 0 and 100. If the -p switch is not\n";
    print "used, the default maximum width of the bar graph is 15%.\n\n";
    print "-r: stores the resulting html and temporary files in a directory. If a\n";
    print "directory is specified then the script creates the directory \"vnAccount\" in\n";
    print "the specified directory. All the specified directories must already exist.\n";
    print "If a directory is not specified then the script attempts to create the directory\n";
    print "\"vnAccount\" in the current directory. If that is not possible then the script\n";
    print "attempts to create the directory \"/tmp/logname/vnAccount\" where \"logname\" is\n"; 
    print "the user name of the current user. If that fails, the script exits. If the\n";
    print "switch is not used the files are created in the current directory.\n\n";
    print "-help: displays this text.\n\n";
    print "Note: this script makes use of a GNU extension. It will run properly only on a\n";
    print "Linux machine, not an SGI.\n\n";
}
#check for valid switches
elsif ($ARGV[0] && $ARGV[0] ne "-r" && $ARGV[0] ne "-p") {
    print "Incorrect switch. Type 'vnAccount -help' to see the switch options\n";
}
else {
    #set defaults
    $normalize = 15;
    $webdirect = ".";
    while (@ARGV) {
	$switch = shift @ARGV;
	next unless ($switch =~ /^-/);
        #check for the directory switch
	if ($switch eq "-r") {
	    #if a directory is given, use that one
	    if ($ARGV[0] && $ARGV[0] =~ /^[^-]/) {
		mkdir("$ARGV[0]/vnAccount", 0777);
		$webdirect = "$ARGV[0]/vnAccount";
	    }
	    else {
		#try to create the directory where vnAccount is called
		$webdirect = "vnAccount";
		mkdir(vnAccount, 0777);
		unless (open(TMEP, "+>vnAccount/tmep")) {
		    #if that fails read the user name of the current user
		    $logname = $ENV{LOGNAME};
		    $webdirect = "/tmp/$logname/vnAccount";
		    #try to create the directories /tmp/$logname and /tmp/$logname/vnAccount
		    mkdir ("/tmp/$logname", 0777); 
		    mkdir("/tmp/$logname/vnAccount", 0777);
		}
		#try to write to the created directory. If that fails, die.
		open (TMEP, "+>$webdirect/tmep") || die "Can't create directory $webdirect : $!\n";
	    }   
	}
	#set maximum value for horizontal scale on bar graph
	elsif ($switch eq "-p") {
	    #check to make sure "-p" is followed by a number between 0 and 100
	    if ($ARGV[0] =~ /\d+/ && $ARGV[0] >= 0 && $ARGV[0] <= 100) {
		$normalize = $ARGV[0];
	    }
	    else {
		die "The switch -p must be followed by a number from 0-100\n";
	    }
	}
    }
    open (TMEP, "+>$webdirect/tmep") || die "Can't create files in directory $webdirect : $!\n";
#sort by time created, oldest first and list in a column
    open(DIRECT, "ls -1 | sort |"); 
    while(<DIRECT>) {
	chop;
	#make sure to pick up only the Rtop files (they start with the number 2)
	if ($_ =~ /^[0-9]{4,4}/) {
	    @lines[$i] = $_;
	    $i = $i + 1;
	}
    }
    
    $k = 0;
#time cluster has been running
    $cluster_time = 0;
    
#go through all the Rtop files, oldest first
    
    open (FIRST, $lines[0]);
    while(<FIRST>) {
#go through all the process lines
	if ($_ =~ /^\d+/) {
	    ($initial, $node, $pid, $user, $pri, $ni, $size, $rss, $share, $stat, $lib, $cpu, $mem, $time, $process) 
= split(/\s+/, $_);
	    #ignore all "idle" or "root" processes
	    unless ($user eq "idle" || $user eq "root" || $node =~ /^[^v]/ || $process eq "top") {
		#distinguish between minutes and minutes:seconds form and turn both into total minutes 
		if ($time =~ /m/) {
		    ($minutes) = split(/m/, $time);
		    $minutes = 2 * $minutes;         #double time since top assumes both processors are working   
		}
		else {
		    ($minutes, $seconds) = split(/:/, $time);
		    $minutes = 2*($minutes + ($seconds / 60));    
		}
		#save the run-time of the first processes so that it can be subtracted off the total later
		$initial_process{$pid} = $minutes . ":" . $user . ":" . $node;
		$initial_time{$user} = ($initial_time{$user} + $minutes);
		
		#create an array which holds the total time used by each user ($total_time)
		#create an array which holds all the individual process data per user
		if ($total_time{$user}) {     #update the arrays 
		    ($usage, $numprocess) = split(/:/, $total_time{$user}); 
		    $$user{$pid} = $node . "+" . $process . "+" . $minutes . "+" . $date_time . "+" . $date_time;
		    $usage = $usage + $minutes;
		    $numprocess++;     #count the number of processes each user has started
		    $total_time{$user} = $usage . ":" . $numprocess;
		}
		else {
		    #create the relevant entries for each array if they don't already exist
		    $$user{$pid} = $node . "+" . $process . "+" . $minutes . "+" . $date_time . "+" . $date_time;
		    $total_time{$user} = $minutes . ":" . 1;
		}
	    }
	}    
#get date stamp on file
	elsif ($_ =~ /^[a-z]/) {
	    chop;
	    ($address, $usage, $date_time) = split(/\s+/, $_, 3);
	    $first_date = $date_time;
	}    
    }
    
    print $k, "\n";
    
#now process the rest of the files
    foreach $j (1..$#lines) {
	$now = $lines[$j];
	#open the current file
	open(NOW, $now) || die "Can't open $now: $!\n";
	while(<NOW>) {
	    #split into pieces
#get all process info
	    if ($_ =~ /^\d+/) {
		($initial, $node, $pid, $user, $pri, $ni, $size, $rss, $share, $stat, $lib, $cpu, $mem, $time, 
		$process) = split(/\s+/, $_);
		#ignore all "idle" or "root" processes
		unless ($user eq "idle" || $user eq "root" || $node =~ /^[^v]/ || $process eq "top") {
		    #distinguish between minutes and minutes:seconds form and turn both into total minutes 
		    if ($time =~ /m/) {
			($minutes) = split(/m/, $time);
			$minutes = 2 * $minutes;         #double time since top assumes both processors are working
		    }
		    else {
			($minutes, $seconds) = split(/:/, $time);
			$minutes = 2*($minutes + ($seconds / 60));
		    }	 
		    #update the total_time and user arrays
		    if ($total_time{$user}) {
			($usage, $numprocess) = split(/:/, $total_time{$user});
			if ($$user{$pid}) {
			    ($master_node, $master_process, $master_time, $master_date, $stop_date) = split(/\+/, 
$$user{$pid}, 5);
			    $$user{$pid} = $master_node. "+" . $master_process . "+" . $minutes . "+" . $master_date
. "+" . $date_time;
			    $usage = $usage + ($minutes - $master_time);
			    $total_time{$user} = $usage . ":" . $numprocess;
			}
			else {
			    $$user{$pid} = $node . "+" . $process . "+" . $minutes . "+" . $date_time . "+" . 
$date_time;
			    $usage = $usage + $minutes;
			    $numprocess++;
			    $total_time{$user} = $usage . ":" . $numprocess;
			}
		    }
		    else { 
			$$user{$pid} = $node . "+" . $process . "+" . $minutes . "+" . $date_time . "+" . $date_time;
			$total_time{$user} = $minutes . ":" . 1;    
		    }
		}
	    }
#get date stamp on file
	    elsif ($_ =~ /^[a-z]/) {
		chop;
		($address, $usage, $date_time) = split(/\s+/, $_, 3);
	    }	
	}
	#print the number of the file that was just read
	print $j, " ";
    }
    print "\n";
#save the date of the oldest file
    $last_date = $date_time;
#print the results to a file
    open(TEMP, "+>$webdirect/temp");
    open(TMEP, "+>$webdirect/tmep");
    
    $all_user = 0;
#subtract the initial run time of processes in the first Rtop files 
    foreach $user (sort keys(%total_time)) {
	($usage, $numprocess) = split(/:/, $total_time{$user});
	if ($initial_time{$user}) {
	    $usage = ($usage - $initial_time{$user});
	    foreach $pid (sort keys(%$user)) {
		next unless ($initial_process{$pid});
		($initial_minutes, $initial_user, $initial_node) = split(/:/, $initial_process{$pid});
		($node, $process, $minutes, $date, $stop_date) = split(/\+/, $$user{$pid}, 5);   
		next unless ($user eq $initial_user && $node eq $initial_node);
		$minutes = ($minutes - $initial_process{$pid});
		$$user{$pid} = $node . "+" . $process . "+" . $minutes . "+" . $date . "+" . $stop_date;
	    }
	}
	$all_user = ($all_user + $usage);
	printf TEMP "%-7f  %-8s  %-5s\n", $usage, $numprocess, $user;
    }
    
    close(TEMP);
    
#sort by time, longest running at the top
    `sort -n -r -o$webdirect/temp $webdirect/temp`;
    
    open(TEMP, "$webdirect/temp");
    open(WEB, "+>$webdirect/data.html");
    
#take the first_date and last_date and calculate the number of minutes between them 
#use the date system call to convert each date to the number of seconds since 00:00:00 Jan 1, 1970
#note: this is a GNU extension to date, it won't work on an SGI 
    $firstdate = `date -d "$first_date" +%s`;
    $lastdate = `date -d "$last_date" +%s`;
    $cluster_time = (($lastdate-$firstdate)/60);
#calculate the total cluster time: the time the cluster has been running multiplied by the number of processors
    $cluster_time = 128*$cluster_time;
#hourmin converts from a time format of minutes to hrs:min format
#the first argument is time in minutes
    &hourmin(*cluster_time, *cluster_hours, *cluster_minutes);
    &hourmin(*all_user, *all_hours, *all_minutes);
    $total_perc = 100*($all_user/$cluster_time);

#start writing HTML code
    printf WEB "<HTML>\n<HEAD>\n<TITLE>Total time used for cluster processes</TITLE>\n</HEAD>\n\n<BODY BGCOLOR = 
\"white\">\n";
    printf WEB "<H1>Summary of <A HREF = \"http://laplace.physics.ubc.ca/People/matt/Doc/VN/index.html\">Cluster</A>
Usage</H1>\n";
    printf WEB "<I>Note: The timings and statistics rooted in this document are currently for informational purposes
only, and are not guaranteed to be completely correct. They do provide, however, a representative view of overall
cluster usage.</I><BR><BR>\n"; 
    printf WEB "<B>Percentage of total cluster time used = %1.1f\%</B><BR>\nTotal cluster time available (hrs : min):
%1s:%02d<BR>\nTotal time used for running processes (hrs : min) = %1s:%02d<BR>\n<BR>\n",$total_perc, $cluster_hours,
$cluster_minutes, $all_hours, $all_minutes;
    printf WEB "The time used by each user is displayed as a percentage of the total cluster time available. The 
width of the bar corresponds to the percentage of time used.\n";
    printf WEB "The height of the bar corresponds to the number of processes run by that user.\n";
    printf WEB "Click on the user name on the right to see each user's process information.<BR>\n<BR>\n";
    printf WEB "Processes counted between <B>%1s</B> and <B>%1s</B>\n", $first_date, $last_date;
#create the main table in the main page
    printf WEB "<TABLE BORDER = \"1\" WIDTH = \"100\%\">\n\t<TR>\n\t<TD><TABLE WIDTH = \"100\%\">\n\t\t<TR> 
<TD ALIGN=\"LEFT\">0</TD>\n\t\t<TD ALIGN=\"RIGHT\">%1s\%&nbsp;</TD></TR>\n\t</TABLE></TD>\n\t<TD>Percentage of time 
used / User / No. processes</TD>\n\t</TR>\n", $normalize;
    
#TEMP is the file that contains all the total_time data sorted by each user's usage time
    while(<TEMP>) {
	chop;
	($user_time,$numprocess,$user) = split(/\s+/, $_);
	$report_percent = 100*($user_time/($cluster_time));    #do the calculations for the lengths of the bar graphs
	$display_percent = 100*($report_percent/$normalize);
	#convert time back to hours:minutes format
	&hourmin(*user_time, *user_hours, *user_minutes);
	#do the calculations for the height of the bar graphs. Height goes as log10 of the number of processes 
        except if number of processes = 1 then height = 1 as well
	if ($numprocess <= 1) {
	    $height = 1;
	}
	else {
	    $height = 30*(log($numprocess)/2.302585093);
	}
	#create a table in each row of the main table. The length of the inner table is set by percent: this creates
the bar graph appearance
	#both the heights of the main table row and the inner table have to be set the same to get the height 
difference
	printf WEB "\t<TR HEIGHT = %1s>\n\t<TD WIDTH = \"70\%\"><TABLE HEIGHT = %1s BORDER = \"0\" BGCOLOR = 
\"black\" WIDTH = \"%1d\%\">\n\t\t<TR><TD><SMALL>&nbsp;</SMALL></TD></TR>\n\t</TABLE></TD>\n\t<TD><A HREF=\"%1s.html\
">%1.1f\%&nbsp;/&nbsp;%1s / %1s</A></TD>\n\t</TR>\n", $height, $height, $display_percent, $user, $report_percent, 
$user, $numprocess;
	open (TEMP1, "+>$webdirect/temp1");
	#print to TEMP1 for sorting by process for each user
	foreach $pid (sort keys(%$user)) {
	    ($master_node, $master_process, $master_time, $master_date, $stop_date) = split(/\+/, $$user{$pid}, 5);
	    printf TEMP1 "%-12s  %-7f  %-5s  %-4s  %-28s+%28s\n", $master_process, $master_time, $pid, $master_node,
$master_date, $stop_date;
	}  
	
	close (TEMP1);
	`sort -o$webdirect/temp1 $webdirect/temp1`;
	open (TEMP1, "$webdirect/temp1");
	open (TEMP2, "+>$webdirect/temp2");
	open (PERSON, "+>$webdirect/$user.html");
	printf PERSON "<HTML>\n<HEAD>\n<TITLE>%ls's process information</TITLE>\n</HEAD>\n\n<BODY>\n", $user;
	printf PERSON "Click on the process to see the usage details for that process\n<BR>\n%1s used %1s:%02d 
(hrs : min) of processing time\n<BR><BR>\n", $user, $user_hours, $user_minutes;
	printf PERSON "<TABLE>\n\t<TR ALIGN = \"LEFT\">\n\t<TH>USER&nbsp;&nbsp;</TH>\n\t<TH>PROCESS&nbsp;&nbsp;</TH>
\n\t<TH>NUMBER OF PROCESSES&nbsp;&nbsp;</TH>\n\t<TH>TIME USED FOR PROCESS&nbsp;&nbsp;</TH>\n\t</TR>\n";
	#initialize lastprocess to something unlikely to be an actual process name
	#numproc counts the number of different processes run by each user; used to create a unique filename for 
PROCESS
	$lastprocess = 1;
	$numproc = 0;
	#TEMP1 is the file that contains all the process data for one user sorted by amount of time used for each 
process
	open (FINAL, "+>$webdirect/final");
	while (<TEMP1>) {
	    chop;
	    ($temp_process, $temp_time, $temp_pid, $temp_node, $temp_date) = split(/\s+/, $_, 5);
	    $nextprocess = $temp_process;
	    #if this is the first line in TEMP1 (no prior process) or if the process hasn't changed then print the 
info to TEMP2 and reassign the lastprocess
	    if ($nextprocess eq $lastprocess || $lastprocess eq 1) {
		printf TEMP2 "%-7f  %5s  %-12s  %-4s  %-56s\n", $temp_time, $temp_pid, $temp_process, $temp_node, 
$temp_date;
		$lastprocess = $nextprocess;	    
	    }
	    #when the process changes: close TEMP2; sort it by time, longest running process first   
	    else {
		#procnum counts the number of times each process was run; printed in PERSON
		$procnum = 0;
		$process_time = 0;
		$numproc++;
		close (TEMP2);
		`sort -n -r -o$webdirect/temp2 $webdirect/temp2`;
		open (TEMP2, "$webdirect/temp2");
		open (PROCESS, "+>$webdirect/$user$numproc");
		printf PROCESS "USER: %-8s\n\n", $user;
		printf PROCESS "       PROCESS       TIME (hrs:min)  PID    NODE  DATE STARTED                  
DATE STOPPED\n";
		#TEMP2 is the file that contains the process data for one process and one user sorted by time
		#open the text file for that process; read the lines from TEMP2 and print to PROCESS
		while (<TEMP2>) {
		    chop;
		    ($master_time, $master_pid, $master_process, $master_node, $master_date) = split(/\s+/, $_, 5);
		    ($start_date, $stop_date) = split(/\+/, $master_date, 2);
		    $process_time = $process_time + $master_time;
		    &hourmin(*master_time, *master_hours, *master_minutes);
		    $procnum++;	
		    if ($stop_date ne $last_date) {
			printf PROCESS "%4s:  %-12s  %11s:%02d  %-5s  %-4s  %-28s  %-28s\n", $procnum, 
$master_process, $master_hours, $master_minutes, $master_pid, $master_node, $start_date, $stop_date;
		    }
		    else {
			printf PROCESS "%4s:  %-12s  %11s:%02d  %-5s  %-4s  %-28s         -\n", $procnum, 
$master_process, $master_hours, $master_minutes, $master_pid, $master_node, $start_date;
		    }
		}
		#update the HTML file for the user which lists the processes
		printf FINAL "%7f  %-8s  %-12s  %2s  %2s\n", $process_time, $user, $master_process, $procnum, 
$numproc;
		close (TEMP2);
		close (PROCESS);
		#open a new TEMP2 and print the new information; reassign lastprocess
		open (TEMP2, "+>$webdirect/temp2");
		printf TEMP2 "%-7f  %5s  %-12s  %-4s  %-28s\n", $temp_time, $temp_pid, $temp_process, $temp_node, 
$temp_date;	    
		$lastprocess = $nextprocess;
	    } 	
	}
	#the last process in TEMP1 won't go through the else section above so now do: sort TEMP2 by time, print 
PROCESS and update PERSON 
	$procnum = 0;
	$process_time = 0;
	$numproc++;
    
	close (TEMP2);
	`sort -n -r -o$webdirect/temp2 $webdirect/temp2`;
	open (TEMP2, "$webdirect/temp2");
	open (PROCESS, "+>$webdirect/$user$numproc");
	printf PROCESS "USER: %-8s\n\n", $user;
	printf PROCESS "       PROCESS       TIME (hrs:min)  PID    NODE  DATE STARTED                  DATE STOPPED
\n";
	while (<TEMP2>) {
	    chop;
	    ($master_time, $master_pid, $master_process, $master_node, $master_date) = split(/\s+/, $_, 5);
	    ($start_date, $stop_date) = split(/\+/, $master_date, 2);
	    $process_time = $process_time + $master_time;
	    &hourmin(*master_time,*master_hours,*master_minutes);
	    $procnum++;
	    if ($stop_date ne $last_date) {
		printf PROCESS "%4s:  %-12s  %11s:%02d  %-5s  %-4s  %-28s  %-28s\n", $procnum, $master_process, 
$master_hours, $master_minutes, $master_pid, $master_node, $start_date, $stop_date;	
	    }
	    else {
		printf PROCESS "%4s:  %-12s  %11s:%02d  %-5s  %-4s  %-28s          -\n", $procnum, $master_process, 
$master_hours, $master_minutes, $master_pid, $master_node, $start_date; 
	    }
	}
	printf FINAL "%7f  %-8s  %-12s  %2s  %2s\n", $process_time, $user, $master_process, $procnum, $numproc;
	close (FINAL);
	`sort -n -r -o$webdirect/final $webdirect/final`;
	close (TEMP2);
	close (PROCESS);
	open (FINAL, "$webdirect/final");
	while (<FINAL>) {
	    chop;
	    ($process_time, $user, $master_process, $procnum, $numproc) = split (/\s+/, $_);
	    &hourmin(*process_time, *process_hours, *process_minutes);
	    printf PERSON "\t<TR ALIGN = \"LEFT\">\n\t<TD>%1s&nbsp;&nbsp;</TD>\n\t<TD><A HREF = \"%1s%1s\">%1s</A>
&nbsp;&nbsp;</TD>\n\t<TD>%1s&nbsp;&nbsp;</TD>\n\t<TD>%1s:%02d&nbsp;&nbsp;</TD>\n\t</TR>\n",$user,$user,$numproc,
$master_process, $procnum, $process_hours, $process_minutes;
	}
	#finish off the HTML for PERSON
	printf PERSON "</TABLE>\n<BR>\n";
	printf PERSON "<A HREF=\"data.html\">Back to total time graph</A>\n</BODY>\n</HTML>\n";
	close (PERSON);
    }
    
#finish off the HTML for WEB
    require "ctime.pl";
    $today = &ctime(time);
    printf WEB "</TABLE><BR>\n<BR>\n<HR>Generated on <I>%1s</I> by <B><A HREF=\"http://laplace.physics.ubc.ca/People/
hkr/vnAccount/vnAccount\">vnAccount</A></B>.<BR>\n", $today;
    printf WEB "Copyright 2000 by <A HREF=\"http://laplace.physics.ubc.ca/People/hkr\">Hanna Ruotsalainen</A>.<BR>
\n";
    printf WEB "Supported by <A HREF=\"http://www.ciar.ca\">CIAR</A> and <A HREF = \"http://www.nserc.ca\">NSERC</A>
<BR>\n";
    printf WEB"</BODY>\n</HTML>";
    print "The files were put in directory ", $webdirect, "\n";
    print "The main page is called data.html\n";
}


#subroutine which changes from minutes format to hrs:min format
sub hourmin {
    #minute_time is passed to the subroutine, hours and minutes are altered in the routine and then accessed back at
the subroutine call  
    local(*minute_time, *hours, *minutes) = @_;
    local ($seconds);
    $hours_float = ($minute_time/60);
    #split on the decimal point
    ($minutes, $seconds) = split(/\./, $minute_time);
    ($hours) = split(/\./, $hours_float);
    #round the seconds up or down to the nearest minute
    if ($seconds > 50) {
	#use mod for the minutes; returns the remainder from minutes/60
	$minutes = ($minutes%60)+1;
	if ($minutes > 59) {
	    $hours++;
	    $minutes = 0;
	}
    }
    else {
	$minutes = ($minutes%60);
    }
}