[imp] IMP/Postgresql extrange problem

[Federico Petronio]

Hello to all:

I use IMP from version 2.0.x. At the moment I have installed version 
2.2.8 with postgresql as database for the "contacts" and sessions. For 
days I have been having the following problem:

IMP screen claims:
"sorry too many clients already"

I known what this means, and it's a postgres matter. But, when I look a 
little deeper I found LOT of this (ps -xa | grep post):

19295 pts/0    S      0:00 /usr/bin/postmaster -D /var/lib/pgsql/data -i 
-B 128
19304 pts/0    S      0:00 postgres: imp_user imp 10.1.1.3 idle
19306 pts/0    D      0:01 postgres: imp_user imp 10.1.1.3 DELETE
19346 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19347 pts/0    S      0:01 postgres: imp_user imp 10.1.1.4 DELETE
19348 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19349 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19350 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19361 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19379 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19385 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19412 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19418 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19419 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19420 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19421 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19422 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 DELETEwaiting
19446 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19465 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19475 pts/0    S      0:00 postgres: imp_user imp 10.1.1.4 idle
19480 pts/0    S      0:00 postgres: imp_user imp 10.1.1.3 idle


10.1.1.4 is the server which have the problem. 10.1.1.3 is a temporal 
server until I find a solution. Both run RH7.2, updated, the same php 
(4.0.6-15) and apache (1.3.22-6). The Postgresql (7.1.3-2) backend is 
always in 10.1.1.4.

I don't understand why everything works fine if I use 10.1.1.3 as the 
web server and 10.1.1.4 as the backend, but the problem appears if I set 
both (web and backend) in 10.1.1.4.

Can it be a matter of insufficient hardware ? Too much work for 10.1.1.4 
when Web and backend are together?

10.1.1.4: Pentium III 550 Mhz, 256Mb RAM, 1 UW-SCSI-2 (almost all the SO 
is here) and 1 IDE HD (with little use). This is a dedicated server, no 
X system.

10.1.1.3: AMD Athlon XP 1500, 1Gb RAM, 1 UW-SCSI-2 and 1 IDE. This is my 
personal PC, with X, netscape, squeak running simulations, etc,etc,etc.

I don't really know about "how much IMP is use" but, if I look in the 
apache log, I found from 3 to 10 lines per second.

An extra note: I have to change the line:

	return (($db = pg_connect($dbconnectstring))) ;
for
	return (($db = pg_pconnect($dbconnectstring))) ;

in horde/imp/lib/db.psgl because strange problems appears with the first 
version of IMP that make the change.

I hope you can help me. Any idea or suggestion is welcome.

Thanks, and sorry about my bad english...
-- 
					Federico Petronio
					petronio@agro.uba.ar
					Linux User #129974

---
Unix IS user friendly. It's just selective about who its friends are.