Re: fcntl(SETLK) [was Re: 2nd update on TOAST] - Mailing list pgsql-hackers

From Alfred Perlstein
Subject Re: fcntl(SETLK) [was Re: 2nd update on TOAST]
Date
Msg-id 20000708144122.G25571@fw.wintelcom.net
Whole thread Raw
In response to Re: fcntl(SETLK) [was Re: 2nd update on TOAST]  (Bruce Momjian <pgman@candle.pha.pa.us>)
List pgsql-hackers
* Bruce Momjian <pgman@candle.pha.pa.us> [000708 06:40] wrote:
> > > That's a good point.  I don't think so because the socket will only
> > > create for one user.  Basically, we don't need something bulletproof
> > > here.  We just need something to prevent admins from accidentally
> > > starting two postmasters on the same port.
> > 
> > Actually I just remebered the issue here, if one wants to start
> > postmaster on an alternate port there will be no conflict and 
> > all hell may break loose.
> 
> We already lock the /data directory.  This is for the port lock.

The whole process could be locked by an fcntl lock on a seperate file,
which I think was already mentioned, however I've deleted most of the
thread unfortunatly.

/tmp/.l.PGSQL.5432 <- fcntl lockfile, aquired first.
/tmp/.s.PGSQL.5432 <- socket.

-- 
-Alfred Perlstein - [bright@wintelcom.net|alfred@freebsd.org]
"I have the heart of a child; I keep it in a jar on my desk."


pgsql-hackers by date:

Previous
From: Noboru Saitou
Date:
Subject: plruby(Re:Trigger function languages)
Next
From: Giles Lean
Date:
Subject: Re: 'errno' undefined?