Wednesday, July 23, 2008

Installing PostgreSQL

This is a quick guide to installing PostgreSQL on a Linux based system. We need to get this up and running before we can write PHP scripts to talk to the database. We need to install PostgreSQL before we can install PHP (since we need the PostgreSQL header files).



This is a little more complicated than installing Apache.



The best bet is to install from packages since this will make things easier, both now and for upgrades.



If you decide to install from source code, you'll need to create a user and compile the source.



Create a PostgreSQL user (you'll need root access). This is who owns the system. I use the "postgres" user and group, you can use something different but it's best to keep it related to the package.



See your system documentation on how to do this (hint - man useradd). Give the user a home directory of /usr/local/pgsql as this is where PostgreSQL defaults to when it installs.



Once all that is done, untar the source:



$ tar -zxf postgresql-.tar.gz (if you got the gzipp'ed version);

or

$ tar -xjf postgresql-.tar.bz2



Jump into the directory:



$ cd postgresql-



(replace with the version you downloaded from the website).



Again, configure has a lot of help:



$ ./configure --help



We'll just install into /usr/local/pgsql and leave the rest as-is:



$ ./configure --prefix=/usr/local/pgsql



Next we have to make it (this will take a while, it takes longer than Apache):



$ make



You will get a line:

All of PostgreSQL successfully made. Ready to install.



when everything is finished.



You'll need to log in as root to do the rest:



$ su -



Go back to the postgresql directory and do:



# make install



PostgreSQL doesn't allow root to start up the server for security reasons, so next you'll have to change the owner of the files:



# chown -R postgres:postgres /usr/local/pgsql



Change to the postgres user:



# su - postgres



Then, change to /usr/local/pgsql:



$ cd /usr/local/pgsql



We need to initialize a database cluster:



$ bin/initdb -D ./data



This will create a /usr/local/pgsql/data directory and initialize it ready for startup and use.



As usual, you can view options by using the --help switch:



$ bin/initdb --help



Then you can start:



$ bin/pg_ctl -D ./data -l data/logfile start



Now that postgresql is started, we need to create a database and another user before we go any further.



You should use a separate database for each of our projects, it's not essential though - it just makes things a little cleaner and easier to understand.



You should also use separate users for each database. This keeps everything separate and 'project a' won't be able to modify any of 'project b's data.



To create a new user in postgresql, it's pretty simple:



$ /usr/local/pgsql/bin/createuser



and follow the prompts - pretty simple :)



The new user should not be able to create new databases or create new users.





To create a database, it's a little different.



$ /usr/local/pgsql/bin/createdb --owner=username databasename



We have to set the owner of the database when we create it, otherwise the 'postgres' user owns it and then we have to grant access to allow our new user to access it.





Now that that's all done, log out of the postgres user, so we're back as root. In the postgres directory, the contrib/start-scripts/ directory has a few system start up scripts.



Follow the instructions in the relevant one to install a start-up script so postgres will start when your system does.



More detailed installation instructions are in the INSTALL file or by going to the PostgreSQL website http://www.postgresql.org/docs/ and going to the right version for you.

0 comments: