[rrd-developers] rrdcached performance with >200k nodes

[Mirek Lauš]

Don't apologize yourself - I do very appreciate your interest.

There is a single batch for each polled hosts, 26 UPDATE lines in average.
Each BATCH is done in separate socket connection, closed with "."

This is the perl code:

sub rrdbatch {
  my ($self, $data) = @_;

  my $cache = new IO::Socket::UNIX( "/tmp/rrdcached.sock");
  $cache->autoflush(0);

  $cache->printflush("BATCH\n");

  for(@$data) {
    $cache->printflush( join(" ", "UPDATE", @$_) . "\n");
  }

  $cache->printflush(".\n");

  $cache->close();
}

-ml

On Wed, Jan 13, 2010 at 8:41 PM, kevin brintnall <kbrint at rufus.net> wrote:
> Hi Mirek,
>
> Forgive the repeated emails..  I am getting ideas one at a time :)
>
> I have another question...   are your "UPDATE" connections long-running?
> Or do you have a new connection for each polling interval (as I do)?
>
> If long running, are you ever closing your BATCH with "." and starting
> anew?
>
> --
>  kevin brintnall =~ /kbrint at rufus.net/
>