On Thu, 24 Feb 2011 03:28:19 +1100, Daniel Manley
<daniel.man...@points.com> wrote:
I've been digging around in my mod_perl-based apps and trying to figure
out why setting up the DB connections for pinging is still randomly
producing first-thing-in-the-morning "mysql server has gone away"
errors. I read up about the morning bug and such (adding ping() to
mysql.pm based on examples on the net) and still can't get it working.
I'm using apache 2.2.17 and mod_perl 2.0.4.
I would suggest looking into your mysql server "wait_timeout".
Calling $dbh->ping() on a handle is not enough to have
seamless recover from stale db connections. At least on MySQL.
Our code, simplifying, looks like:
----------------- 8< -----------------------------------
sub is_alive {
my ($dbh) = @_;
return if not $dbh;
if (not eval { $dbh->FETCH('Active') && $dbh->ping() }) {
warn "Found stale dbh handle: $dbh ($@)\n";
return;
}
return 1;
}
sub db_connect {
my $dbh = cached_dbh();
return $dbh if is_alive($dbh);
$dbh = new_db_connection();
return $dbh;
}
----------------- 8< -----------------------------------
--
Cosimo
