How do I find the path to pg_hba.conf from the shell?
I would like to obtain the path to pg_hba.conf from the shell. The path varies between versions of PostgreSQL. For instance, for 8.4 and 9.1:
/etc/postgresql/8.4/main/pg_hba.conf
/etc/postgresql/9.1/main/pg_hba.conf
I have tried the pg_config command, but it does not seem to include this information.
This is so that I can use a uniform command for opening pg_hba.conf and other PostgreSQL configuration files for editing.
I'm using bash.
pg_config is for compliation information, to help extensions and client programs compile and link against PostgreSQL. It knows nothing about the active PostgreSQL instance(s) on the machine, only the binaries.
pg_hba.conf can appear in many other places depending on how Pg was installed. The standard location is pg_hba.conf within the data_directory of the database (which could be in /home, /var/lib/pgsql, /var/lib/postgresql/[version]/, /opt/postgres/, etc etc etc) but users and packagers can put it wherever they like. Unfortunately.
The only valid ways find pg_hba.conf is to ask a running PostgreSQL instance where it's pg_hba.conf is, or ask the sysadmin where it is. You can't even rely on asking where the datadir is and parsing postgresql.conf because an init script might passed a param like -c hba_file=/some/other/path when starting Pg.
What you want to do is ask PostgreSQL:
SHOW hba_file;
This command must be run on a superuser session, so for shell scripting you might write something like:
psql -t -P format=unaligned -c 'show hba_file';
and set the environment variables PGUSER, PGDATABASE, etc to ensure that the connection is right.
Yes, this is somewhat of a chicken-and-egg problem, in that if the user can't connect (say, after screwing up editing pg_hba.conf) you can't find pg_hba.conf in order to fix it.
Another option is to look at the ps command's output and see if the postmaster data directory argument -D is visible there, e.g.
ps aux | grep 'postgres *-D'
since pg_hba.conf will be inside the data directory (unless you're on Debian/Ubuntu or some derivative and using their packages).
If you're targeting specifically Ubuntu systems with PostgreSQL installed from Debian/Ubuntu packages it gets a little easier. You don't have to deal with hand-compiled-from-source Pg that someone's initdb'd a datadir for in their home dir, or an EnterpriseDB Pg install in /opt, etc. You can ask pg_wrapper, the Debian/Ubuntu multi-version Pg manager, where PostgreSQL is using the pg_lsclusters command from pg_wrapper.
If you can't connect (Pg isn't running, or you need to edit pg_hba.conf to connect) you'll have to search the system for pg_hba.conf files. On Mac and Linux something like sudo find / -type f -name pg_hba.conf will do. Then check the PG_VERSION file in the same directory to make sure it's the right PostgreSQL version if you have more than one. (If pg_hba.conf is in /etc/, ignore this, it's the parent directory name instead). If you have more than one data directory for the same PostgreSQL version you'll have to look at database size, check the command line of the running postgres from ps to see if it's data directory -D argument matches where you're editing, etc.
This is how I do it:
john@host:~$ ls /etc/postgresql
9.1
john@host:~$ sudo vim /etc/postgresql/9.1/main/pg_hba.conf
Since the path is like this: /etc/postgresql/[VERSION]/main/pg_hba.conf
I am using the following for config files detection:
$(ls /etc/postgresql/*/main/pg_hba.conf)
$(ls /etc/postgresql/*/main/postgresql.conf)
Using command-line, ask your database:
$ psql -U postgres
postgres=# SHOW config_file;
- Copied from Stackoverflow - https://stackoverflow.com/a/3603162/5677857
Editing the correct pg_hba.conf with your default visual editor (vim, emacs, nano, joe, etc.) is as easy as:
$VISUAL /etc/postgresql/$(ls /etc/postgresql)/main/pg_hba.conf