Postgres Installation
For a fresh install or upgrading from previous releases of Postgres:
Read any last minute information and platform specific porting notes. There are some platform specific notes at the end of this file for Ultrix4.x, Linux, BSD/OS and NeXT. There are other files in directory /usr/src/pgsql/doc, including files FAQ-Irix and FAQ-Linux. Also look in directory ftp://ftp.postgresql.org/pub. If there is a file called INSTALL in this directory then this file will contain the latest installation information.
Please note that a "tested" platform in the list given earlier simply means that someone went to the effort at some point of making sure that a Postgres distribution would compile and run on this platform without modifying the code. Since the current developers will not have access to all of these platforms, some of them may not compile cleanly and pass the regression tests in the current release due to minor problems. Any such known problems and their solutions will be posted in ftp://ftp.postgresql.org/pub/INSTALL.
Create the Postgres superuser account (postgres is commonly used) if it does not already exist.
The owner of the Postgres files can be any unprivileged user account. It must not be root, bin, or any other account with special access rights, as that would create a security risk.
Log in to the Postgres superuser account. Most of the remaining steps in the installation will happen in this account.
Ftp file ftp://ftp.postgresql.org/pub/postgresql-v6.5.1.tar.gz from the Internet. Store it in your home directory.
Some platforms use flex. If your system uses flex then make sure you have a good version. To check, type
$ flex --version
If the flex command is not found then you probably do not need it. If the version is 2.5.2 or 2.5.4 or greater then you are okay. If it is 2.5.3 or before 2.5.2 then you will have to upgrade flex. You may get it at ftp://prep.ai.mit.edu/pub/gnu/flex-2.5.4.tar.gz.
If you need flex and don't have it or have the wrong version, then you will be told so when you attempt to compile the program. Feel free to skip this step if you aren't sure you need it. If you do need it then you will be told to install/upgrade flex when you try to compile Postgres.
You may want to do the entire flex installation from the root account, though that is not absolutely necessary. Assuming that you want the installation to place files in the usual default areas, type the following:
$ su - $ cd /usr/local/src ftp prep.ai.mit.edu ftp> cd /pub/gnu/ ftp> binary ftp> get flex-2.5.4.tar.gz ftp> quit $ gunzip -c flex-2.5.4.tar.gz | tar xvf - $ cd flex-2.5.4 $ configure --prefix=/usr $ gmake $ gmake check # You must be root when typing the next line: $ gmake install $ cd /usr/local/src $ rm -rf flex-2.5.4
This will update files /usr/man/man1/flex.1, /usr/bin/flex, /usr/lib/libfl.a, /usr/include/FlexLexer.h and will add a link /usr/bin/flex++ which points to flex.
If you are not upgrading an existing system then skip to step 9. If you are upgrading from 6.5, you do not need to dump/reload or initdb. Simply compile the source code, stop the postmaster, do a "make install", and restart the postmaster. If you are upgrading from 6.4.* or earlier, back up your database. For alpha- and beta-level releases, the database format is liable to change, often every few weeks, with no notice besides a quick comment in the HACKERS mailing list. Full releases always require a dump/reload from previous releases. It is therefore a bad idea to skip this step.
Tip: Do not use the pg_dumpall script from v6.0 or everything will be owned by the Postgres super user.
To dump your fairly recent post-v6.0 database installation, type
$ pg_dumpall > db.out
To use the latest pg_dumpall script on your existing older database before upgrading Postgres, pull the most recent version of pg_dumpall from the new distribution:
$ cd $ gunzip -c postgresql-v6.5.1.tar.gz \ | tar xvf - src/bin/pg_dump/pg_dumpall $ chmod a+x src/bin/pg_dump/pg_dumpall $ src/bin/pg_dump/pg_dumpall > db.out $ rm -rf src
If you wish to preserve object id's (oids), then use the -o option when running pg_dumpall. However, unless you have a special reason for doing this (such as using OIDs as keys in tables), don't do it.
If the pg_dumpall command seems to take a long time and you think it might have died, then, from another terminal, type
$ ls -l db.outseveral times to see if the size of the file is growing.
Please note that if you are upgrading from a version prior to Postgres95 v1.09 then you must back up your database, install Postgres95 v1.09, restore your database, then back it up again. You should also read the release notes which should cover any release-specific issues.
Caution |
You must make sure that your database is not updated in the middle of your backup. If necessary, bring down postmaster, edit the permissions in file /usr/local/pgsql/data/pg_hba.conf to allow only you on, then bring postmaster back up. |
If you are upgrading an existing system then kill the postmaster. Type
$ ps -ax | grep postmasterThis should list the process numbers for a number of processes. Type the following line, with pid replaced by the process id for process postmaster. (Do not use the id for process "grep postmaster".) Type
$ kill pidto actually stop the process.
Tip: On systems which have Postgres started at boot time, there is probably a startup file which will accomplish the same thing. For example, on my Linux system I can type
$ /etc/rc.d/init.d/postgres.init stopto halt Postgres.
If you are upgrading an existing system then move the old directories out of the way. If you are short of disk space then you may have to back up and delete the directories instead. If you do this, save the old database in the /usr/local/pgsql/data directory tree. At a minimum, save file /usr/local/pgsql/data/pg_hba.conf.
Type the following:
$ su - $ cd /usr/src $ mv pgsql pgsql_6_0 $ cd /usr/local $ mv pgsql pgsql_6_0 $ exit
If you are not using /usr/local/pgsql/data as your data directory (check to see if environment variable PGDATA is set to something else) then you will also want to move this directory in the same manner.
Make new source and install directories. The actual paths can be different for your installation but you must be consistent throughout this procedure.
Note: There are two places in this installation procedure where you will have an opportunity to specify installation locations for programs, libraries, documentation, and other files. Usually it is sufficient to specify these at the gmake install stage of installation.
Type
$ su $ cd /usr/src $ mkdir pgsql $ chown postgres:postgres pgsql $ cd /usr/local $ mkdir pgsql $ chown postgres:postgres pgsql $ exit
Unzip and untar the new source file. Type
$ cd /usr/src/pgsql $ gunzip -c ~/postgresql-v6.5.1.tar.gz | tar xvf -
Configure the source code for your system. It is this step at which you can specify your actual installation path for the build process (see the --prefix option below). Type
$ cd /usr/src/pgsql/src $ ./configure [ options ]
Among other chores, the configure script selects a system-specific "template" file from the files provided in the template subdirectory. If it cannot guess which one to use for your system, it will say so and exit. In that case you'll need to figure out which one to use and run configure again, this time giving the --with-template=TEMPLATE option to make the right file be chosen.
Please Report Problems: If your system is not automatically recognized by configure and you have to do this, please send email to scrappy@hub.org with the output of the program ./config.guess. Indicate what the template file should be.
Choose configuration options. Check Configuration Options for details. However, for a plain-vanilla first installation with no extra options like multi-byte character support or locale collation support it may be adequate to have chosen the installation areas and to run configure without extra options specified. The configure script accepts many additional options that you can use if you don't like the default configuration. To see them all, type
./configure --helpSome of the more commonly used ones are:
--prefix=BASEDIR Selects a different base directory for the
installation of the Postgres configuration.
The default is /usr/local/pgsql.
--with-template=TEMPLATE
Use template file TEMPLATE - the template
files are assumed to be in the directory
src/template, so look there for proper values.
--with-tcl Build interface libraries and programs requiring
Tcl/Tk, including libpgtcl, pgtclsh, and pgtksh.
--with-perl Build the Perl interface library.
--with-odbc Build the ODBC driver package.
--enable-hba Enables Host Based Authentication (DEFAULT)
--disable-hba Disables Host Based Authentication
--enable-locale Enables USE_LOCALE
--enable-cassert Enables ASSERT_CHECKING
--with-CC=compiler
Use a specific C compiler that the configure
script cannot find.
--with-CXX=compiler
--without-CXX
Use a specific C++ compiler that the configure
script cannot find, or exclude C++ compilation
altogether. (This only affects libpq++ at
present.)
Here is the configure script used on a Sparc Solaris 2.5 system with /opt/postgres specified as the installation base directory:
$ ./configure --prefix=/opt/postgres \ --with-template=sparc_solaris-gcc --with-pgport=5432 \ --enable-hba --disable-locale
Tip: Of course, you may type these three lines all on the same line.
Install the man and HTML documentation. Type
$ cd /usr/src/pgsql/doc $ gmake install
The documentation is also available in Postscript format. Look for files ending with .ps.gz in the same directory.
Compile the program. Type
$ cd /usr/src/pgsql/src $ gmake all >& make.log & $ tail -f make.log
The last line displayed will hopefully be
All of PostgreSQL is successfully made. Ready to install.Remember, “gmake” may be called “make” on your system. At this point, or earlier if you wish, type control-C to get out of tail. (If you have problems later on you may wish to examine file make.log for warning and error messages.)
Note: You will probably find a number of warning messages in make.log. Unless you have problems later on, these messages may be safely ignored.
If the compiler fails with a message stating that the flex command cannot be found then install flex as described earlier. Next, change directory back to this directory, type
$ gmake cleanthen recompile again.
Compiler options, such as optimization and debugging, may be specified on the command line using the COPT variable. For example, typing
$ gmake COPT="-g" all >& make.log &would invoke your compiler's -g option in all steps of the build. See src/Makefile.global.in for further details.
Install the program. Type
$ cd /usr/src/pgsql/src $ gmake install >& make.install.log & $ tail -f make.install.log
The last line displayed will be
gmake[1]: Leaving directory `/usr/src/pgsql/src/man'At this point, or earlier if you wish, type control-C to get out of tail. Remember, “gmake” may be called “make” on your system.
If necessary, tell your system how to find the new shared libraries. You can do one of the following, preferably the first:
As root, edit file /etc/ld.so.conf. Add a line
/usr/local/pgsql/libto the file. Then run command /sbin/ldconfig.
In a bash shell, type
export LD_LIBRARY_PATH=/usr/local/pgsql/lib
In a csh shell, type
setenv LD_LIBRARY_PATH /usr/local/pgsql/lib
Please note that the above commands may vary wildly for different operating systems. Check the platform specific notes, such as those for Ultrix4.x or and for non-ELF Linux.
If, when you create the database, you get the message
pg_id: can't load library 'libpq.so'then the above step was necessary. Simply do this step, then try to create the database again.
If you used the --with-perl option to configure, check the install log to see whether the Perl module was actually installed. If you've followed our advice to make the Postgres files be owned by an unprivileged userid, then the Perl module won't have been installed, for lack of write privileges on the Perl library directories. You can complete its installation, either now or later, by becoming the user that does own the Perl library (often root) (via su) and doing
$ cd /usr/src/pgsql/src/interfaces/perl5 $ gmake install
If it has not already been done, then prepare account postgres for using Postgres. Any account that will use Postgres must be similarly prepared.
There are several ways to influence the runtime environment of the Postgres server. Refer to the Administrator's Guide for more information.
Note: The following instructions are for a bash/sh shell. Adapt accordingly for other shells.
Add the following lines to your login environment: shell, ~/.bash_profile:
PATH=$PATH:/usr/local/pgsql/bin MANPATH=$MANPATH:/usr/local/pgsql/man PGLIB=/usr/local/pgsql/lib PGDATA=/usr/local/pgsql/data export PATH MANPATH PGLIB PGDATA
Several regression tests could fail if the user's locale collation scheme is different from that of the standard C locale.
If you configure and compile Postgres with --enable-locale then you should set the locale environment to “C” (or unset all “LC_*” variables) by putting these additional lines to your login environment before starting postmaster:
LC_COLLATE=C LC_CTYPE=C export LC_COLLATE LC_CTYPE
Make sure that you have defined these variables before continuing with the remaining steps. The easiest way to do this is to type:
$ source ~/.bash_profile
Create the database installation from your Postgres superuser account (typically account postgres). Do not do the following as root! This would be a major security hole. Type
$ initdb
Set up permissions to access the database system. Do this by editing file /usr/local/pgsql/data/pg_hba.conf. The instructions are included in the file. (If your database is not located in the default location, i.e. if PGDATA is set to point elsewhere, then the location of this file will change accordingly.) This file should be made read only again once you are finished. If you are upgrading from v6.0 or later you can copy file pg_hba.conf from your old database on top of the one in your new database, rather than redoing the file from scratch.
Briefly test that the backend will start and run by running it from the command line.
Start the postmaster daemon running in the background by typing
$ cd $ nohup postmaster -i > pgserver.log 2>&1 &
Create a database by typing
$ createdb
Connect to the new database:
$ psql
And run a sample query:
postgres=> SELECT datetime 'now';
Exit psql:
postgres=> \q
Remove the test database (unless you will want to use it later for other tests):
$ destroydb
Run postmaster in the background from your Postgres superuser account (typically account postgres). Do not run postmaster from the root account!
Usually, you will want to modify your computer so that it will automatically start postmaster whenever it boots. It is not required; the Postgres server can be run successfully from non-privileged accounts without root intervention.
Here are some suggestions on how to do this, contributed by various users.
Whatever you do, postmaster must be run by the Postgres superuser (postgres?) and not by root. This is why all of the examples below start by switching user (su) to postgres. These commands also take into account the fact that environment variables like PATH and PGDATA may not be set properly. The examples are as follows. Use them with extreme caution.
If you are installing from a non-privileged account and have no root access, then start the postmaster and send it to the background:
$ cd $ nohup postmaster > regress.log 2>&1 &
Edit file rc.local on NetBSD or file rc2.d on SPARC Solaris 2.5.1 to contain the following single line:
su postgres -c "/usr/local/pgsql/bin/postmaster -S -D /usr/local/pgsql/data"
In FreeBSD 2.2-RELEASE edit /usr/local/etc/rc.d/pgsql.sh to contain the following lines and make it chmod 755 and chown root:bin.
#!/bin/sh [ -x /usr/local/pgsql/bin/postmaster ] && { su -l pgsql -c 'exec /usr/local/pgsql/bin/postmaster -D/usr/local/pgsql/data -S -o -F > /usr/local/pgsql/errlog' & echo -n ' pgsql' }You may put the line breaks as shown above. The shell is smart enough to keep parsing beyond end-of-line if there is an expression unfinished. The exec saves one layer of shell under the postmaster process so the parent is init.
In RedHat Linux add a file /etc/rc.d/init.d/postgres.init which is based on the example in contrib/linux/. Then make a softlink to this file from /etc/rc.d/rc5.d/S98postgres.init.
In RedHat Linux edit file /etc/inittab to add the following as a single line:
pg:2345:respawn:/bin/su - postgres -c "/usr/local/pgsql/bin/postmaster -D/usr/local/pgsql/data >> /usr/local/pgsql/server.log 2>&1 </dev/null"(The author of this example says this example will revive the postmaster if it dies, but he doesn't know if there are other side effects.)
Run the regression tests. The file /usr/src/pgsql/src/test/regress/README has detailed instructions for running and interpreting the regression tests. A short version follows here:
Type
$ cd /usr/src/pgsql/src/test/regress $ gmake clean $ gmake all runtest
You do not need to type gmake clean if this is the first time you are running the tests.
You should get on the screen (and also written to file ./regress.out) a series of statements stating which tests passed and which tests failed. Please note that it can be normal for some tests to "fail" on some platforms. The script says a test has failed if there is any difference at all between the actual output of the test and the expected output. Thus, tests may "fail" due to minor differences in wording of error messages, small differences in floating-point roundoff, etc, between your system and the regression test reference platform. "Failures" of this type do not indicate a problem with Postgres. The file ./regression.diffs contains the textual differences between the actual test output on your machine and the "expected" output (which is simply what the reference system produced). You should carefully examine each difference listed to see whether it appears to be a significant issue.
For example,
For a i686/Linux-ELF platform, no tests failed since this is the v6.5.1 regression testing reference platform.
Even if a test result clearly indicates a real failure, it may be a localized problem that will not affect you. An example is that the int8 test will fail, producing obviously incorrect output, if your machine and C compiler do not provide a 64-bit integer data type (or if they do but configure didn't discover it). This is not something to worry about unless you need to store 64-bit integers.
Conclusion? If you do see failures, try to understand the nature of the differences and then decide if those differences will affect your intended use of Postgres. The regression tests are a helpful tool, but they may require some study to be useful.
After running the regression tests, type
$ destroydb regression $ cd /usr/src/pgsql/src/test/regress $ gmake cleanto recover the disk space used for the tests. (You may want to save the regression.diffs file in another place before doing this.)
If you haven't already done so, this would be a good time to modify your computer to do regular maintainence. The following should be done at regular intervals:
Minimal Backup Procedure
Run the SQL command VACUUM. This will clean up your database.
Back up your system. (You should probably keep the last few backups on hand.) Preferably, no one else should be using the system at the time.
Ideally, the above tasks should be done by a shell script that is run nightly or weekly by cron. Look at the man page for crontab for a starting point on how to do this. (If you do it, please e-mail us a copy of your shell script. We would like to set up our own systems to do this too.)
If you are upgrading an existing system then reinstall your old database. Type
$ cd $ psql -e template1 < db.outIf your pre-v6.2 database uses either path or polygon geometric data types, then you will need to upgrade any columns containing those types. To do so, type (from within psql)
UPDATE FirstTable SET PathCol = UpgradePath(PathCol); UPDATE SecondTable SET PathCol = UpgradePath(PathCol); ... VACUUM;UpgradePath() checks to see that a path value is consistant with the old syntax, and will not update a column which fails that examination. UpgradePoly() cannot verify that a polygon is in fact from an old syntax, but RevertPoly() is provided to reverse the effects of a mis-applied upgrade.
If you are a new user, you may wish to play with Postgres as described below.
Clean up after yourself. Type
$ rm -rf /usr/src/pgsql_6_5 $ rm -rf /usr/local/pgsql_6_5 # Also delete old database directory tree if it is not in # /usr/local/pgsql_6_5/data $ rm ~/postgresql-v6.5.1.tar.gz
You will probably want to print out the documentation. If you have a Postscript printer, or have your machine already set up to accept Postscript files using a print filter, then to print the User's Guide simply type
$ cd /usr/local/pgsql/doc $ gunzip user.ps.tz | lpr
Here is how you might do it if you have Ghostscript on your system and are writing to a laserjet printer.
$ alias gshp='gs -sDEVICE=laserjet -r300 -dNOPAUSE' $ export GS_LIB=/usr/share/ghostscript:/usr/share/ghostscript/fonts $ gunzip user.ps.gz $ gshp -sOUTPUTFILE=user.hp user.ps $ gzip user.ps $ lpr -l -s -r manpage.hp
The Postgres team wants to keep Postgres working on all of the supported platforms. We therefore ask you to let us know if you did or did not get Postgres to work on you system. Please send a mail message to pgsql-ports@postgresql.org telling us the following:
The version of Postgres (v6.5.1, 6.5, beta 990318, etc.).
Your operating system (i.e. RedHat v5.1 Linux v2.0.34).
Your hardware (SPARC, i486, etc.).
Did you compile, install and run the regression tests cleanly? If not, what source code did you change (i.e. patches you applied, changes you made, etc.), what tests failed, etc. It is normal to get many warning when you compile. You do not need to report these.
Now create, access and manipulate databases as desired. Write client programs to access the database server. In other words, enjoy!