freebsd stable - Jan 2006 - postgresql doesn't start on boot-up

I'm new to FreeBSD and for the first time ever, I cvsupp'ed  
to freebsd-stable 6 and rebuilt everthing, following the instructions 
in the handbook. It went pretty smoothly except that several services 
failed to start and others acted badly.

In particular:

Initially, postfix, cups and postgresql failed to start.

named started too late, so ntpd couldn't resolve the timeserver names
and couldn't set the time. I solved this per the suggestion in a prior 
post by setting:

early_late_divider="NETWORKING" in rc.conf.

cups didn't start because it didn't understand the faststart argument
it was being passed. I just recompiled Samba without CUPS since I don't 
have any printers anyway.

The one I can't figure out is postgresql. It doesn't start, but after
the server finishes booting, I can start it manually by running the
010pgsql.sh script.

Also, this is probably a stupid question, but is there a boot.log or
something like that shows the bootup messages that come after dmesg? 

-- 
Ken Stevenson
Allen-Myland Inc.

Ken Stevenson wrote:
> I'm new to FreeBSD and for the first time ever, I cvsupp'ed  
> to freebsd-stable 6 and rebuilt everthing, following the instructions 
> in the handbook. It went pretty smoothly except that several services 
> failed to start and others acted badly.
> 
> In particular:
> 
> Initially, postfix, cups and postgresql failed to start.
This sounds like you don't have the most recent versions of these ports 
installed, as all of them have had fixes for these problems. It's also 
possible that you have stale files in /usr/local/etc/rc.d that are not 
related to ports that you have installed. For example, a lot of users have 
at some time in the past copied cups.sh.sample to cups.sh, which prevents 
updates to the script from being seen. For old style scripts it's safer to 
symlink the foo.sh.sample file to foo.sh, fyi.
> named started too late, so ntpd couldn't resolve the timeserver names
> and couldn't set the time. 
This is a more interesting problem. On all of my test systems, named is the 
first to run after SERVERS, followed immediately by ntpdate and ntpd. There 
is an argument to be made that it would actually be better to run ntp* 
first, using IP addresses in its config files, and then run named, since if 
you're running a name server for profit (as opposed to fun) accurate 
timestamps on the logs are very valuable to diagnose startup problems. OTOH, 
it's reasonable to assume for the general case that the current default 
order is good, so I have no objection to adding named to the REQUIRE: line 
in ntpdate. Ken, can you give this a try and let us know how it works for 
you? I'll hold off on making this change till others have a chance to
comment.
> I solved this per the suggestion in a prior 
> post by setting:
> 
> early_late_divider="NETWORKING" in rc.conf.
Ummm ... are these systems running in a jail? Because that's the only time 
that the solution you mentioned should really be used (or make a difference).
> Also, this is probably a stupid question, but is there a boot.log or
> something like that shows the bootup messages that come after dmesg? 
In /etc/syslog.conf, uncomment the line that mentions console.log, 'touch 
/var/log/console.log', and then reboot after you make the change I suggested
above.

hth,

Doug

-- 

     This .signature sanitized for your protection