[rrd-developers] [PATCH] examples/rrdcached

[kevin brintnall]

This patch includes a couple support files for RRDCached:

 * script to estimate rrdcached's memory size and I/O utilization
 * example "BATCH" mode client

---
diff --git a/examples/rrdcached/RRDCached.pm b/examples/rrdcached/RRDCached.pm
new file mode 100644
index 0000000..d776e32
--- /dev/null
+++ b/examples/rrdcached/RRDCached.pm
@@ -0,0 +1,90 @@
+
+use strict;
+use warnings;
+
+package RRDCached;
+
+=head1 RRDCached
+
+This module implements the B<RRDCached> client protocol for bulk updates.
+
+=head1 SYNOPSIS
+
+    my $cache = RRDCached->new('unix:/var/run/rrdcached.sock')
+        or die "Cannot connect to RRDCached";
+
+    $cache->update('file1.rrd', 'N:10:2:78');
+    $cache->update('file2.rrd', '1222973760:30:0:9', 'N:68:1:55');
+    ...
+
+    $cache->done();
+
+=cut
+
+use IO::Socket;
+
+#################################################################
+
+sub new {
+    my ($class, $daemon) = @_;
+    my $this = {};
+
+    $daemon ||= $ENV{RRDCACHED_ADDRESS};
+    defined $daemon or return undef;
+
+    my $sock_family = "INET";
+
+    if ($daemon =~ m{^unix: | ^/ }x)
+    {
+        $sock_family = "UNIX";
+        $daemon =~ s/^unix://;
+    }
+
+    my $sock = "IO::Socket::$sock_family"->new($daemon)
+        or die "Cannot connect to daemon";
+
+    $sock->printflush("BATCH\n");
+
+    my $go = $sock->getline;
+    warn "We didn't get go-ahead from rrdcached" unless $go =~ /^0/;
+
+    $sock->autoflush(0);
+
+    bless { sock => $sock,
+            daemon => $daemon,
+        }, $class;
+}
+
+sub update {
+    my $this = shift;
+    my $file = shift;
+    ## @updates = @_;
+
+    @_ or warn "No updates for $file!";
+
+    ## rrdcached doesn't handle N: timestamps
+    my $now = time();
+    s/^N(?=:)/$now/ for (@_);
+
+    $this->{sock}->print("update $file @_\n");
+}
+
+sub done {
+    my ($this) = @_;
+
+    my $sock = delete $this->{sock};
+
+    $sock->printflush(".\n");
+    my $errs = $sock->getline;
+
+    my ($num_err) = $errs =~ /^(\d+)/;
+    return unless $num_err;
+
+    $sock->getline for (1..$num_err);
+
+    $sock->close;
+}
+
+#################################################################
+
+1;
diff --git a/examples/rrdcached/rrdcached-size.pl b/examples/rrdcached/rrdcached-size.pl
new file mode 100755
index 0000000..df6aaf6
--- /dev/null
+++ b/examples/rrdcached/rrdcached-size.pl
@@ -0,0 +1,153 @@
+#!/usr/bin/perl
+
+=head1 NAME
+
+rrdcached-size.pl - estimate the IO and memory requirements for rrdcached
+
+=head1 SYNOPSIS
+
+B<rrdcached-size.pl>
+[B<-rrds>E<nbsp>I<file_count>]
+[B<-step>E<nbsp>I<seconds>]
+[B<-update>E<nbsp>I<length>]
+[B<-file>E<nbsp>I<length>]
+[B<-io>E<nbsp>I<files/sec>]
+[B<-w>E<nbsp>I<seconds>]
+[B<-f>E<nbsp>I<seconds>]
+[B<-pagesize>E<nbsp>I<bytes>]
+
+=head1 OPTIONS
+
+=over 4
+
+=item B<-rrds> I<file_count>
+
+Specify the number of RRDs in the working set.
+
+=item B<-step> I<seconds>
+
+Specify the RRD step value for each file.
+
+=item B<-update> I<length>
+
+Average update string length.  For this calculation, the time value must
+be specified as a C<time_t>, not C<N>.  For example, this update string
+would lead to B<-update>E<nbsp>I<43> :
+
+  1226936851:0:0:101113914:0:0:0:25814373:0:0
+
+=item B<-file> I<length>
+
+Specify the average file name length.  For this calculation, use the full
+path of the file.
+
+=item B<-io> I<files/sec>
+
+Specify the number of RRD files that your system can write per second.
+
+=item B<-w> I<timer>
+
+Specifies the B<-w> timer used with rrdcached.  For more information, see
+the B<rrdcached> documentation.
+
+=item B<-f> I<timer>
+
+Specifies the B<-f> timer used with rrdcached.  For more information, see
+the B<rrdcached> documentation.
+
+=item B<-pagesize> I<bytes>
+
+Manually specify the system page size, in case it is not detected
+properly.
+
+=back
+
+=cut
+
+use strict;
+use warnings;
+
+my $filename_len = 60;
+my $update_len = 128;
+my $rrds = 100;
+my $step = 300;
+my $rrd_per_sec = 200;
+my $rrdc_write = 300;
+my $rrdc_flush = 3600;
+my $pagesize = `pagesize` || 4096;
+
+#################################################################
+
+use Getopt::Long;
+GetOptions('rrds=i' => \$rrds,
+           'step=i' => \$step,
+           'update=i' => \$update_len,
+           'file=i' => \$filename_len,
+           'io=i' => \$rrd_per_sec,
+           'w=i'    => \$rrdc_write,
+           'f=i'    => \$rrdc_flush,
+           'pagesize=i' => \$pagesize,
+           'h' => \&usage,
+           )
+    or die "Options failure";
+
+ at ARGV and die "Extra args: @ARGV\n";
+
+#################################################################
+
+my $MEG = 1024*1024;
+
+my $write_time = int($rrds / $rrd_per_sec);
+my $write_busy = int(100 * $write_time / $rrdc_write);
+my $buffered_pdp = $rrdc_write / $step;
+
+my $max_ram
+    = $rrds
+    * ($filename_len
+           + ( $rrdc_write / $step ) * $update_len)
+    / $MEG;
+
+my $journal_size
+    = $rrds
+    * (length("update") + $filename_len + $update_len + 3)
+    * ($rrdc_flush/$step)
+    * 2  # 2 logs
+    / $MEG;
+
+my $journal_rate = (($journal_size*$MEG/2))/$rrdc_flush;
+my $journal_page_rate = $journal_rate / $pagesize;
+
+$_ = sprintf("%.1f", $_)
+    for ($write_time,
+         $write_busy,
+         $buffered_pdp,
+         $max_ram,
+         $journal_size,
+         $journal_rate,
+         $journal_page_rate,
+     );
+
+print <<"EOF";
+RRD files     : $rrds files
+RRD step      : $step seconds
+Update length : $update_len bytes
+IO writes/sec : $rrd_per_sec rrd/sec
+write timer   : $rrdc_write seconds
+flush timer   : $rrdc_flush seconds
+-----------------------------------------------------------------
+
+Time to write all RRDs: $write_time sec ($write_busy\% busy)
+
+$buffered_pdp PDPs will be buffered per file
+
+RAM usage: $max_ram MB
+
+Journal size: $journal_size MB (total size for two journals)
+
+Journal write rate: $journal_page_rate page/sec ($journal_rate byte/sec)
+EOF
+
+sub usage {
+    system("perldoc $0");
+    exit(1);
+}