On Fri, Jun 25, 2010 at 08:22:35AM -0600, Serge Dubrouski wrote:
> On Fri, Jun 25, 2010 at 5:56 AM, Lars Ellenberg
> <[email protected]> wrote:
> > On Fri, Jun 25, 2010 at 01:49:32PM +0200, Dejan Muhamedagic wrote:
> >> On Fri, Jun 25, 2010 at 12:58:15PM +0200, Lars Ellenberg wrote:
> >> > On Thu, Jun 24, 2010 at 11:59:55AM -0600, Serge Dubrouski wrote:
> >> > > Attached is an improved patch.
> >> > >
> >> > > 1. It parses unix_socket_directory out from PostgreSQL configuration
> >> > > file and used its value as a default value of OCF_RESKEY_socketdir.
> >> > > (Perl is used for that)
> >> >
> >> > sed -n  -e 's/#.*$//; s/[[:space:]]*$//;' \
> >> >     -e 's/^[[:space:]]*'"$parameter"'[[:space:]=]\+[[:space:]]*//p;' 
> >> > postgresql.conf |
> >> >     tail -n1
> >> > should do it, too.
> >>
> >> Isn't then the last s command enough?
> >
> > First strip optional inline comments and possibly trailing whitespace.
> > Yes, you could put it in one s///, with s/...()(|)/\1/p, but
> > that won't get more readable.
> >
> >> I believe that it was in case the directory was in /var/run/
> >> which is cleaned on boot.
> >
> > Bad choice, then ;-)
> > Choose differently, or re-create it on boot.
> > Ah well, or patch the resource agent.
> > Whatever works best.
> 
> It's not a users choice I'm afraid. Changing that parameter in
> postgresql.conf from it's default value creates more problems for DBA
> than gives advantages since psql tool doesn't tolerate such change and
> one have to provide the "non-default" value as a command line
> parameter. So as I said earlier I think that Ubuntu developers for
> some reason made that choice for their users. Now our user will have
> to take care of that bu setting additional OCF parameter since. Since
> it's a default value for that distro its value isn't explicitly set in
> postgresql.conf and our parsing mechanism won't help.

FUBAR :(
any way to ask postgres about its hardcoded defaults?
sort of like "postconf" for postfix?

Still I don't think it belongs into the RA.
This is needed once after reboot,
not for every start/stop/monitor of the RA.
Rather put a line into rc.local,
or into /etc/default/pacemaker (yes, that's shell code,
and it is sourced from the heartbeat and corosync init scripts).

-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
_______________________________________________________
Linux-HA-Dev: [email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/

Reply via email to