PostgreSQL 8.0.3 Documentation - Chapter 23. Monitoring Database Activity

Am Dienstag, 26. Juli 2005 12:06 schrieb Halley Pacheco de Oliveira:

The user, database, and connection source host items remain the same for
the life of the client connection, but the activity indicator changes.

So the database name didn't remain the same, neither the process id.
Connecting to another database creates a brand new connection using other
server process as it seems to be. I think it should be written in this
paragraph. I couldn't imagine that, for me the process was still the same.

What is the problem here? You create a new connection, so you get the
information in the new connection, as documented.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

There is no problem, just a surprise. Once I was connected to one database I didn't expect that a
connection to another database inside psql would close one process and fork a new process. Once
the chapter is about Monitoring Database Activity, I think that it would be good to alert the
readers about this fact. Maybe other users dont't expect this fact too. Just that.

Regards,
Halley

--- Peter Eisentraut <peter_e@gmx.net> escreveu:

Am Dienstag, 26. Juli 2005 12:06 schrieb Halley Pacheco de Oliveira:

The user, database, and connection source host items remain the same for
the life of the client connection, but the activity indicator changes.

So the database name didn't remain the same, neither the process id.
Connecting to another database creates a brand new connection using other
server process as it seems to be. I think it should be written in this
paragraph. I couldn't imagine that, for me the process was still the same.

What is the problem here? You create a new connection, so you get the
information in the new connection, as documented.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

_______________________________________________________
Yahoo! Acesso Gr�tis - Internet r�pida e gr�tis.
Instale o discador agora! http://br.acesso.yahoo.com/

# halleypo@yahoo.com.br / 2005-07-26 13:48:55 +0000:

--- Peter Eisentraut <peter_e@gmx.net> escreveu:

Am Dienstag, 26. Juli 2005 12:06 schrieb Halley Pacheco de Oliveira:

The user, database, and connection source host items remain the same for
the life of the client connection, but the activity indicator changes.

So the database name didn't remain the same, neither the process id.
Connecting to another database creates a brand new connection using other
server process as it seems to be. I think it should be written in this
paragraph. I couldn't imagine that, for me the process was still the same.

What is the problem here? You create a new connection, so you get the
information in the new connection, as documented.

There is no problem, just a surprise. Once I was connected to one database I didn't expect that a
connection to another database inside psql would close one process and fork a new process. Once
the chapter is about Monitoring Database Activity, I think that it would be good to alert the
readers about this fact. Maybe other users dont't expect this fact too. Just that.

If anywhere, such information would belong to the psql(1)
documentation, where it already is:

\connect (or \c) [ dbname [ username ] ]
Establishes a connection to a new database and/or under a user
name. The previous connection is closed.

--
How many Vietnam vets does it take to screw in a light bulb?
You don't know, man. You don't KNOW.
Cause you weren't THERE. http://bash.org/?255991

What surprises me is not that the previous connection is closed and a new connection is opened.
What surprises me is the fact that the previous process is killed and a new process is forked just
to connect to another database with the same user. I was expecting that the new connection could
use the same process. If you are monitoring user activity you must know that when the user
connects to another database another process is created, or even worst, if the user connects to
the same database again another process is created.

psql -U sistemas -h 192.168.0.1 teste

ps aux | grep ^postgres
postgres 3910 0.0 0.3 20320 2000 ? S 10:31 0:00 /usr/bin/postmaster -p 5432 -D
/var/lib/pgsql/data
postgres 3968 0.0 0.3 11120 1792 ? S 10:31 0:00 postgres: stats buffer process
postgres 3969 0.0 0.3 10128 1844 ? S 10:31 0:00 postgres: stats collector process
postgres 14006 0.0 0.5 20720 2640 ? S 14:08 0:00 postgres: sistemas teste
192.168.0.1 idle

teste=# \c teste

ps aux | grep ^postgres
postgres 3910 0.0 0.3 20320 2000 ? S 10:31 0:00 /usr/bin/postmaster -p 5432 -D
/var/lib/pgsql/data
postgres 3968 0.0 0.3 11120 1792 ? S 10:31 0:00 postgres: stats buffer process
postgres 3969 0.0 0.3 10128 1844 ? S 10:31 0:00 postgres: stats collector process
postgres 14038 0.0 0.5 20720 2608 ? S 14:09 0:00 postgres: sistemas teste
192.168.0.1 idle

It was not clear to me this relationship between connection and process. Now it is.

Regards,
Halley

--- Roman Neuhauser <neuhauser@sigpipe.cz> escreveu:

If anywhere, such information would belong to the psql(1)
documentation, where it already is:

\connect (or \c) [ dbname [ username ] ]
Establishes a connection to a new database and/or under a user
name. The previous connection is closed.

_______________________________________________________
Yahoo! Acesso Gr�tis - Internet r�pida e gr�tis.
Instale o discador agora! http://br.acesso.yahoo.com/

# halleypo@yahoo.com.br / 2005-07-26 14:17:17 -0300:

What surprises me is not that the previous connection is closed and a
new connection is opened.

What surprises me is the fact that the previous process is killed and
a new process is forked just to connect to another database with the
same user. I was expecting that the new connection could use the same
process. If you are monitoring user activity you must know that when
the user connects to another database another process is created, or
even worst, if the user connects to the same database again another
process is created.

It was not clear to me this relationship between connection and
process. Now it is.

Well, there's nothing that would make this model required, but it
surely helps keep things like entropy at reasonable levels.
IOW, this behavior allows for greater stability through simpler code
and limited process lifetime.

But maybe I'm not understanding the point you're trying to make.
To make things a bit clearer: what is it that you find so disturbing
or surprising in the current PostgreSQL behavior? Why did you expect
it reusing the same process, and what benefits do you expect (or
preferably, have experimentally gained) from the alternative?

--
How many Vietnam vets does it take to screw in a light bulb?
You don't know, man. You don't KNOW.
Cause you weren't THERE. http://bash.org/?255991

The surprise is:

Oracle - MTS - Multi-Threaded-Server - MTS allows many user processes to share very few server
processes. Without MTS, each user process requires its own dedicated server process; a new server
process is created for each client requesting a connection. A dedicated server process remains
associated to the user process for the remainder of the connection. With MTS many user processes
connect to a dispatcher process. The dispatcher routes client requests to the next available
shared server process. The advantage of MTS is that system overhead is reduced, so the number of
users that can be supported is increased.

Contrasting with this in PostgreSQL a new process is forked just to connect to another database.
The PostgreSQL behavior seems similar to old Oracle versions.

Regards,
Halley
--- Roman Neuhauser <neuhauser@sigpipe.cz> escreveu:

Well, there's nothing that would make this model required, but it
surely helps keep things like entropy at reasonable levels.
IOW, this behavior allows for greater stability through simpler code
and limited process lifetime.

But maybe I'm not understanding the point you're trying to make.
To make things a bit clearer: what is it that you find so disturbing
or surprising in the current PostgreSQL behavior? Why did you expect
it reusing the same process, and what benefits do you expect (or
preferably, have experimentally gained) from the alternative?

_______________________________________________________
Yahoo! Acesso Gr�tis - Internet r�pida e gr�tis.
Instale o discador agora! http://br.acesso.yahoo.com/

# halleypo@yahoo.com.br / 2005-07-26 22:17:12 -0300:

--- Roman Neuhauser <neuhauser@sigpipe.cz> escreveu:

But maybe I'm not understanding the point you're trying to make.
To make things a bit clearer: what is it that you find so disturbing
or surprising in the current PostgreSQL behavior? Why did you expect
it reusing the same process, and what benefits do you expect (or
preferably, have experimentally gained) from the alternative?

The surprise is:

Oracle - MTS - Multi-Threaded-Server - MTS allows many user processes
to share very few server processes. Without MTS, each user process
requires its own dedicated server process; a new server process is
created for each client requesting a connection. A dedicated server
process remains associated to the user process for the remainder of
the connection. With MTS many user processes connect to a dispatcher
process. The dispatcher routes client requests to the next available
shared server process. The advantage of MTS is that system overhead is
reduced, so the number of users that can be supported is increased.

I came to PostgreSQL from MySQL, which is multithreaded, and found
PostgreSQL absolutely UNsurprising in this aspect. There's quite
a few populare programs that behave similarly.

Contrasting with this in PostgreSQL a new process is forked just to
connect to another database. The PostgreSQL behavior seems similar to
old Oracle versions.

So what? Is this a troll?

--
How many Vietnam vets does it take to screw in a light bulb?
You don't know, man. You don't KNOW.
Cause you weren't THERE. http://bash.org/?255991

On Tue, Jul 26, 2005 at 10:17:12PM -0300, Halley Pacheco de Oliveira wrote:

The surprise is:

Oracle - MTS - Multi-Threaded-Server - MTS allows many user processes
to share very few server processes. Without MTS, each user process
requires its own dedicated server process; a new server process is
created for each client requesting a connection. A dedicated server
process remains associated to the user process for the remainder of
the connection. With MTS many user processes connect to a dispatcher
process. The dispatcher routes client requests to the next available
shared server process. The advantage of MTS is that system overhead is
reduced, so the number of users that can be supported is increased.

Hmm, seems a very similar thing can be had in Postgres by means of
pgpool.

It'd be interesting to see how are cursors handled, or any other
long-lived session-local resources for that matter, by this dispatcher
process of Oracle's.

--
Alvaro Herrera (<alvherre[a]alvh.no-ip.org>)
"Crear es tan dif�cil como ser libre" (Elsa Triolet)

--- Alvaro Herrera <alvherre@alvh.no-ip.org> escreveu:

Hmm, seems a very similar thing can be had in Postgres by means of
pgpool.

You're right. Using a JDBC connection pool to cache database connections, reuse physical
connections and minimize expensive operations in the creation and closing of database sessions
(1), or pgpool that is a connection pool server for PostgreSQL (2), the overhead of create a
server process for each new connection almost disappears.

It will only be necessary one connection pool for each database accessed, and pgpool will make a
new connection if there's no user name and database name pair yet.

Once more the middle tier simplifies the database operation.

Regards,
Halley

(1)
http://www.oracle.com/technology/sample_code/tech/java/sqlj_jdbc/files/9i_jdbc/OCIMidAuthSample/Readme.html

(2) http://pgpool.projects.postgresql.org/

__________________________________________________
Converse com seus amigos em tempo real com o Yahoo! Messenger
http://br.download.yahoo.com/messenger/