[Pacemaker] postgresql monitoring- "incomplete startup packet"- getting filled in log

Joe Healy joehealy at gmail.com
Fri Feb 19 07:40:34 EST 2010

Hopefully not going too far of topic, the code (in perl) that is doing this

# Check whether a postmaster server is running at the specified port.
# Arguments: <version> <cluster> <port>
sub cluster_port_running {
    die "port_running: invalid port $_[2]" if $_[2] !~ /\d+/;
    my $socketdir = get_cluster_socketdir $_[0], $_[1];
    my $socketpath = "$socketdir/.s.PGSQL.$_[2]";
    return 0 unless -S $socketpath;

    socket(SRV, PF_UNIX, SOCK_STREAM, 0) or die "socket: $!";
    my $running = connect(SRV, sockaddr_un($socketpath));
    close SRV;
    return $running ? 1 : 0;

I think it raises a couple of questions about using this init.d script for
pacemaker. The debian maintainers of postgresql have done a wonderful job of
pulling together a system for running multiple versions of postgresql
together on the one host. This means the init.d script may not do exactly
what you want in this case.

I don't know exactly, but you should check what the return code of
/etc/init.d/postgresql status is in various scenarios (including more than
one version installed at once). Possibly your maintenance procedures will
prevent this case occuring in production though.

It may also be the case that for proper pacemaker status checking, you want
a more thorough check of the db being up than just the socket being open.

Hope that helps.


On Fri, Feb 19, 2010 at 9:28 PM, Andrew Beekhof <andrew at beekhof.net> wrote:

> On Fri, Feb 19, 2010 at 9:22 AM, Jayakrishnan <jayakrishnanlll at gmail.com>
> wrote:
> > But the issues is that my postgresql log file is getting filled
> > with
> >
> > 2010-02-19 13:19:30 IST LOG:  incomplete startup packet
> > 2010-02-19 13:19:45 IST LOG:  incomplete startup packet
> > 2010-02-19 13:20:00 IST LOG:  incomplete startup packet
> > ......
> >
> > messages like this. I googled and I fount that my pgsql monitor is is the
> > cause for this. You can see that the message is given at the same instant
> as
> > the monitor interval.
> >
> > Is there any way to get rid of this??
> Perhaps asking on the postgres list might get you a more informed response
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20100219/ccf4316f/attachment-0002.html>

More information about the Pacemaker mailing list