git-svn-id: https://ceph.svn.sf.net/svnroot/ceph@1643 29311d96-e01e-0410-9327-a35deaa...

diff --git a/branches/sage/mds/script/study_find.pl b/branches/sage/mds/script/study_find.pl
new file mode 100755 (executable)
index 0000000..7e4c760
--- /dev/null
+++ b/branches/sage/mds/script/study_find.pl
@@ -0,0 +1,172 @@
+#!/usr/bin/perl
+
+use strict;
+
+my $name = shift @ARGV;
+my $dsfn = shift @ARGV;
+my $fnlenfn = shift @ARGV;
+my $logfn = shift @ARGV;
+
+my $nfiles = 0;
+my $ndirs = 0;
+my $nreg = 0;
+my $nhardlinks = 0;
+my %nlinks;
+my %names;
+my %dirsize;
+my %fnlen;  
+my $nfnchars;
+
+my $bytes;
+my $ebytes;
+
+#
+# output generated with
+#
+#  find . -path ./.snapshot -prune -o -exec ls -dilsn --time-style=+%s \{\} \;
+#
+# find output looks like this:
+#4495744 4 drwxrwxrwx  24 0 0 4096 1187290970 .
+#2996320 8 drwxr-xr-x  189 0 1000 8192 1186594257 ./jangle
+#28378499 4 drwxr-x--x  4 1068885 52673 4096 1162938122 ./jangle/cymcruise
+#28378500 4 drwx--S---  5 1068885 52673 4096 1162938122 ./jangle/cymcruise/Maildir
+#28378501 4 drwx------  2 1068885 52673 4096 1162938122 ./jangle/cymcruise/Maildir/tmp
+#28378502 4 drwx------  2 1068885 52673 4096 1162938122 ./jangle/cymcruise/Maildir/new
+#28378503 4 drwx------  2 1068885 52673 4096 1162938122 ./jangle/cymcruise/Maildir/cur
+#28378504 4 -rw-r--r--  1 1068885 52673 260 943743700 ./jangle/cymcruise/.alias
+#999425 4 drwxr-xr-x  92 1125 100 4096 1186523060 .
+#999426 0 lrwxrwxrwx  1 0 0 5 1177701093 ./root -> /root
+#1015809 4 drwxr-xr-x  4 1289 1000 4096 1174584949 ./andrea
+#541007 4 drwxr-xr-x  3 0 0 4096 1173111449 ./andrea/lux
+#5014055 4 drwx--S---  11 70228 51207 4096 1172250346 ./andrea/lux/Maildir
+
+# dirs we're currently counting in
+my %numindir;
+
+sub finish_dir {
+    my $curdir = shift @_;
+    #print "finish_dir $curdir\n";
+    $dirsize{$numindir{$curdir}}++;
+    delete $numindir{$curdir};
+}
+
+my $curdir;
+$ndirs = 1;
+while (<>) {
+    #print;
+    chomp;
+    my ($ino, $blah, $mode, $blah, $nlink, $uid, $gid, $size, $mtime, @path) = split(/ /,$_);
+    my $file = join(' ',@path);
+    ($file) = split(/ \-\> /, $file);
+    my @bits = split(/\//, $file);
+    my $f = pop @bits;
+    my $dir = join('/', @bits);
+    #print "file = '$file', dir = '$dir', curdir = '$curdir'\n";
+
+    if ($dir ne $curdir) {
+       for my $d (keys %numindir) {
+           #print "? $d vs $dir\n";
+           &finish_dir($d) if ($d ne substr($dir, 0, length($d)));
+       }
+       $curdir = $dir;
+    }
+
+    my $esize = 0;
+    $esize = int (($size-1)/4096)*4096 + 4096 if $size > 0;
+    $esize += 160;  # for the inode?
+    $bytes += $size;
+    $ebytes += $esize;
+
+    $nfiles++; 
+    $numindir{$dir}++;
+
+    my $fnlen = length($f);
+    $fnlen{$fnlen}++;
+    $nfnchars += $fnlen;
+    
+    if ($mode =~ /^d/) {
+       $ndirs++;
+       # find does depth-first search, so assume we descend, so that on empty dir we "back out" above and &finish_dir.
+       $numindir{$file} = 0;
+       $curdir = $file;
+    } else {
+       $nreg++ if $mode =~ /^f/;
+       if ($nlink > 1) {
+           #system "ls -aldi $file";
+           $nhardlinks++;
+           $nlinks{$nlink}++;
+           push(@{$names{$ino}->{$dir}}, $file);
+       }
+    }
+}
+for my $d (keys %numindir) {
+    &finish_dir($d);
+}
+
+
+
+my $nsamedir = 0;
+open(LOG, ">$logfn") if $logfn;
+for my $ino (keys %names) {
+    print LOG "# $ino\n";
+    my @dirs = keys %{$names{$ino}};
+    my $insamedir = 1 if scalar(@dirs) == 1;
+    for my $dir (@dirs) {
+       print LOG "#\t$dir\n";
+       for my $fn (@{$names{$ino}->{$dir}}) {
+           print LOG "#\t\t$fn\n";
+           $nsamedir++ if $insamedir;
+       }
+    }
+}
+close LOG;
+
+sub do_cdf {
+    my $hash = shift @_;
+    my $num = shift @_;
+    my $sum = shift @_;
+    my $fn = shift @_;
+
+    open(CDF, ">$fn") if $fn;
+    print CDF "# $name\n# val\tnum\n";
+
+    my $median;
+    my $p = 0;
+    for my $v (sort {$a <=> $b} keys %$hash) {
+       print CDF "$v\t$hash->{$v}\n";
+       $p += $hash->{$v} * $v;
+       if (!(defined $median) &&
+           $p >= ($sum/2)) {
+           $median = $v;
+       }
+    }
+    my $avg = sprintf("%.2f", $sum/$num);
+    print CDF "# avg $avg, median $median, sum $sum, num $num\n";
+    return ($avg, $median);
+}
+close DSLOG;
+
+
+my ($avgdirsize, $mediandirsize) = &do_cdf(\%dirsize, $ndirs, $nfiles, $dsfn);
+my ($avgfnlen, $medianfnlen) = &do_cdf(\%fnlen, $nfiles, $nfnchars, $fnlenfn);
+
+
+# stat fs
+#my $df = `df $base`;
+#my $line = (split(/\n/,$df))[1]; # second line
+#my ($kb) = $df =~ /\s+\d+\s+(\d+)/;
+my $gb = sprintf("%.1f",($ebytes / 1024 / 1024 / 1024));
+
+# final line
+my $pad = '# ' . (' ' x (length($name)-2));
+print "$pad\tgb\tfiles\tdirs\tdsavg\tdsmed\tfnavg\tfnmed\treg\tnl>1\tsmdr\tnlink=2\t=3\t=4\t...\n";
+print "$name\t$gb\t$nfiles\t$ndirs\t$avgdirsize\t$mediandirsize\t$avgfnlen\t$medianfnlen\t$nreg\t$nhardlinks\t$nsamedir";
+my $i = 2;
+for (sort {$a <=> $b} keys %nlinks) {
+    while ($_ < $i) {
+       print "\t0";
+    }
+    print "\t$nlinks{$_}";
+    $i = $_ + 1;
+}
+print "\n";