Re: Bugs in PostgreSQL v7.1 rpms for RedHat

David Lizano wrote:
> I have installed PostgreSQL v7.1 in RedHat from RPM's and I have find some
> bugs:

> Stopping the service:
> ===============
> When '/etc/rc.d/init.d/postgresql stop' try to stop 'postmaster' can't find
> the 'postmaster.pid' file, and don't stop the postmaster daemon.

> Y think it try to get in from $PGDATA/postmaster.pid , and it is locate in
> '/var/run/postmaster.pid', so it don't works.

Works here, for some reason. But I'll check it. I run these very RPM's
on RH 6.2, 7.0, _and_ 7.1 and have not found this problem. But, I _will_
double check, and will issue a patch that still uses pg_ctl accordingly.

> In the stop zone add:
> killall -TERM postmaster

I would _not_ do this. Use pg_ctl, as it is designed for the job. If a
parameter needs to be changed, or a symlink made, then that's fine. But
there are problems using killall for this purpose.

> logs
> ===
> The logs goes to /dev/null.

Intentionally.

> change "pg_ctl > /dev/null"
>
> by
> "pg_ctl > var/log/postgresql"

If you want rollable logging, enable syslog and friends in
postgresql.conf, add the right line in syslog.conf, and restart
postmaster. A redirected log such as the above will require a postmaster
shutdown in order to roll. But, if you want to do that, it does work.

> Access 'postmaster' through TCP/IP (it is not formerly a bug)
> ============================================
> By default yo can't acces through TCP/IP, add this:
> -o '-i'
> then:
>
> pg_ctl -o '-i'

Use the postgresql.conf file instead. This is intentional -- call it a
bug if you wish, but it is an intentional change, and it is documented
as such in README.rpm-dist, in the section entitled 'Grand Unified
Configuration File', beginning on line 296. I am shipping the default
postgresql.conf.sample, and am not patching it in any way shape or form,
so that the main documentation will at least be correct here. Maybe I
should have made the announcement state the change a little more
clearly, though: but I figured people would read the documentation I
took time to write.

> Starting "PostgreSQL" when the server does.
> ================================
> The installation script don't do this, you must do it if you wants that
> 'postmaster' starts when the server does.

> ln -s /etc/rc.d/init.d/postgresql /etc/rc.d/rc3.d/S50postgresql
<snip>
> ln -s /etc/rc.d/init.d/postgresql /etc/rc.d/rc6.d/K25postgresql

Use /sbin/chkconfig --add postgresql--it's easier. The right information
is set in the initscript for chkconfig usage. This used to be
automatically done as part of the %post installation scriptlet -- I
figured it's best to let the admin handle this, rather than make it
automatic. It _is_ documented in README.rpm-dist, starting at line
number 276.

While README.rpm-dist might be a little hard to find on some systems,
use 'rpm -ql postgresql | grep README\.rpm-dist' to find it, as it is
part of the main RPM.

Thanks much for taking the time to report these things.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11