Thread: What could cause postmaster to be started many times?

I ran into a situation where (without any known changes) a simple login
script would fail ("sorry buddy, session headers already sent") and each
time it seemed a new postmaster would be started. ps -aux showed about
8+ processes this morning but I find that hard to believe because the
sandbox machine only has 256 MB RAM and load was low (in my haste, I did
not note all information). There are no scripts that can start the
postmaster. I just use /usr/local/pgsql/bin/postmaster -S -i -D
/usr/local/pgsql/data. I am not sure of the cause and effect, or how
things got better.

What could cause this kind of behavior and how might I avoid it in the
future? Any clues?

PG 7.0.3 on FreeBDS 4.1 and  php 4.0.4pl1. There were no extra
.s.PGSQL.5432 sockets.

Thanks, Jeff

"Jeff Fitzmyers" <jeff.fitzmyers@managestar.com> writes:
> I ran into a situation where (without any known changes) a simple login
> script would fail ("sorry buddy, session headers already sent") and each
> time it seemed a new postmaster would be started.

Are you sure these are postmasters and not backends?

            regards, tom lane

Not sure, (I'm a newbie) but how long would the processes last? I am not
using persistant pg_pconnect (although I think php is configured to alow
them). The 8ish I saw today lasted at least a minute based on a login
scrip that only has 1 query on a table of 40 rows.

After testing with this in mind, I can see the postmaster backend
(thanks for the education!) appear and persist until the query (I guess)
finishes - which is awhile because the browser window is waiting for
something.

From your comment it seems the db is working perfectly. I will look more
into the php scripts, Thank you very much, Jeff

"Jeff Fitzmyers" <jeff.fitzmyers@managestar.com> writes:
> I ran into a situation where (without any known changes) a simple
login
> script would fail ("sorry buddy, session headers already sent") and
each
> time it seemed a new postmaster would be started.

Are you sure these are postmasters and not backends?

            regards, tom lane