On Wed, Feb 14, 2007 at 10:38:40PM +0100, Peter Kovacs wrote: > The timeout part is clear. But what about the process ownership part? > Is it not that the story has two parts with the second part being > about making sure that (in case postgres has not been [could not be] > shut down properly) the postmaster recognizes, at startup, that the > pid file it has found is stale?
The process ownership part seems to be about the startup script for Linux. I looked at this: http://www.faqs.org/docs/ppbook/c16573.htm and what I see leads me to think that starting the postmaster process through a normal SMF service start method using the method_context to run it as "postgres" is good enough. See the "Using pg_ctl" and "Using the SysV Script" sections. Well, SMF is not like a SysV init system. AFAICS the SMF start method for PostgreSQL should use pg_ctl as per the "Using pg_ctl" section. Nico --
