Is Postgres broken in Red Hat 9?

On Tue, 8 Jul 2003, Dragan Matic wrote:

Hi, there is a small problem here I have, I would appreciate any
suggestion.

We were using Postgres 7.2.x and later 7.3.x under Red Hat 8, and
everything worked fine. Clients are working under windows and are
communicating to Postgres via ODBC. This weekend we tried to upgrade to
Red Hat 9 (and PG 7.3.3 shipped with it), but we just couldn't connect
to server through ODBC. Postgres starts fine, I can access it through
psql, but we simply can't access it from client machines. It is
configured properly (it should accept tcp/ip connections, and
appropriate rights are given). I simply can't find what is going wrong.
I even trien putting older (Postgres made) rpms, but the problem
persists. Are there any similar experiences, is this a RedHat bug, has
it closed something somehow, or where should I start looking? Tnx.

Can you connect to it locally by using your machine's IP address? I.e. if
your machine's eth0 sits on 10.0.0.2, does psql -h 10.0.0.2 work?

You may have a firewall setup to block all ports by default.
firewall-config or something like it was the name of the firewall config
util in 7.2.

On Tuesday 08 Jul 2003 6:38 pm, Dragan Matic wrote:

Hi, there is a small problem here I have, I would appreciate any
suggestion.

We were using Postgres 7.2.x and later 7.3.x under Red Hat 8, and
everything worked fine. Clients are working under windows and are
communicating to Postgres via ODBC. This weekend we tried to upgrade to
Red Hat 9 (and PG 7.3.3 shipped with it), but we just couldn't connect
to server through ODBC. Postgres starts fine, I can access it through
psql, but we simply can't access it from client machines. It is
configured properly (it should accept tcp/ip connections, and
appropriate rights are given). I simply can't find what is going wrong.
I even trien putting older (Postgres made) rpms, but the problem
persists. Are there any similar experiences, is this a RedHat bug, has
it closed something somehow, or where should I start looking? Tnx.

1. As root, try the following to see if PG is listening:
lsof -i | grep postgres
You should see a postgres process with a TCP connection in (LISTEN) mode.

2. Try the following from another one of your PCs to make sure there's not a
firewall in the way (open a DOS-prompt)
telnet myPGmachine:5432
Hit return twice and you should see an error message like:
EFATAL: invalid length of startup packet

3. Turn on the logging in the ODBC driver settings and see if there's anything
useful there. If nothing leaps out at you, try back here.

HTH
--
Richard Huxton

On Tuesday 08 Jul 2003 6:38 pm, Dragan Matic wrote:

We were using Postgres 7.2.x and later 7.3.x under Red Hat 8, and
everything worked fine. Clients are working under windows and are
communicating to Postgres via ODBC.

Oh - one thing you need to know - there was an extra column introduced to the
pg_hba.conf file, I think that might have been between 7.2 and 7.3

--
Richard Huxton

On Tue, 8 Jul 2003 11:40:15 -0600 (MDT) "scott.marlowe" <scott.marlowe@ihs.com> wrote:

You may have a firewall setup to block all ports by default.
firewall-config or something like it was the name of the firewall config
util in 7.2.

in recent redhats, this is likely it.

sudo /sbin/ipchains -L

to see the current ipchains firewall rules,

sudo /sbin/iptables -L

for iptables. i've not run RH 9 yet, so i don't know what the defaults are
here, but i've had this problem on slightly older releases, so there are no
suprises if this is the problem.

you can either learn enough about iptables/ipchains to punch a hole, but to
make sure that it will solve the problem, try shutting the whole think down
first:

sudo /etc/rc.d/init.d/ipchains stop
sudo /etc/rc.d/init.d/iptables stop

after that, see if the network clients can connect.

richard
--
Richard Welty rwelty@averillpark.net
Averill Park Networking 518-573-7592
Java, PHP, PostgreSQL, Unix, Linux, IP Network Engineering, Security

Tnx everybody for help, it wasn't any problem with blocked ports, but in
ODBC driver.

This is from the Postgres Site:

"In addition, this driver fixes an annoying problem with recent builds
of PostgreSQL on Redhat systems that have a particularly long compiler
version string that caused a buffer overflow resulting in a GPF on
Windows upon connection to the database. "

Dragan Matic wrote:

On Wednesday 09 July 2003 09:00, Dragan Matic wrote:

This is from the Postgres Site:

"In addition, this driver fixes an annoying problem with recent builds
of PostgreSQL on Redhat systems that have a particularly long compiler
version string that caused a buffer overflow resulting in a GPF on
Windows upon connection to the database. "

This information should be updated, as this problem probably would occur on
any system using gcc 3.x, as that is the package with the long version
string. It shouldn't be a Red Hat specific bug.

This issue should have been present on Red Hat 8.0 as well, since it's
virtually the same compiler. IIRC, you said you went from RH 7.3 up. 7.3
used the much-maligned Red Hat-specific gcc 2.96, whose version string is
much shorter. Test it yourself with 'gcc --version' and see what you get.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11