Can not connect to local PostgreSQL

I've managed to bork my local development environment.

All my local Rails apps are now giving the error:

PGError
could not connect to server: Permission denied
	Is the server running locally and accepting
	connections on Unix domain socket "/var/pgsql_socket/.s.PGSQL.5432"?

I've no idea what's caused this.

While searching for a solution I've updated all bundled gems, updated system gems, updated MacPorts. No joy.

Others have reported this issue when upgrading from OSX Leopard to Lion, due to confusion over which version of Postgres should be used (i.e., OSX version or MacPorts version). I've been running Lion for several months, so it seems strange that this should happen now.

I'm reluctant to mess around too much without first understanding what the problem is. How can I debug this methodically?

How can I determine how many versions of PostgreSQL are on my system, which one is being accessed, and where it is located? How do I fix this if the wrong PostgreSQL is being used?

Sorry for the noob questions. I'm still learning how this works! Thanks for any pointers.

EDIT

Some updates based on suggestions and comments below.

I tried to run pg_lsclusters which returned a command not found error.

I then tried to local my pg_hba.conf file and found these three sample files:

/opt/local/share/postgresql84/pg_hba.conf.sample
/opt/local/var/macports/software/postgresql84/8.4.7_0/opt/local/share/postgresql84/pg_hba.conf.sample
/usr/share/postgresql/pg_hba.conf.sample

So I assume 3 versions of PSQL are installed? Macports, OSX default and ???.

I then did a search for the launchctl startup script ps -ef | grep postgres which returned

0    56     1   0 11:41AM ??         0:00.02 /opt/local/bin/daemondo --label=postgresql84-server --start-cmd /opt/local/etc/LaunchDaemons/org.macports.postgresql84-server/postgresql84-server.wrapper start ; --stop-cmd /opt/local/etc/LaunchDaemons/org.macports.postgresql84-server/postgresql84-server.wrapper stop ; --restart-cmd /opt/local/etc/LaunchDaemons/org.macports.postgresql84-server/postgresql84-server.wrapper restart ; --pid=none
  500   372     1   0 11:42AM ??         0:00.17 /opt/local/lib/postgresql84/bin/postgres -D /opt/local/var/db/postgresql84/defaultdb
  500   766   372   0 11:43AM ??         0:00.37 postgres: writer process                                                                                                                                                                                                                                                                                                                   
  500   767   372   0 11:43AM ??         0:00.24 postgres: wal writer process                                                                                                                                                                                                                                                                                                               
  500   768   372   0 11:43AM ??         0:00.16 postgres: autovacuum launcher process                                                                                                                                                                                                                                                                                                      
  500   769   372   0 11:43AM ??         0:00.08 postgres: stats collector process                                                                                                                                                                                                                                                                                                          
  501  4497  1016   0 12:36PM ttys000    0:00.00 grep postgres

I've posted the contents of postgresql84-server.wrapper at http://pastebin.com/Gj5TpP62.

I tried to run port load postgresql184-server but received an error Error: Port postgresql184-server not found.

I'm still very confused how to fix this, and appreciate any "for dummies" pointers.

Thanks!

EDIT2

This issue began after I had some problems with daemondo. My local Rails apps were crashing with an application error along the lines of "daemondo gem can not be found". I then went through a series of bundle updates, gem updates, port updates and brew updates to try and find the issue.

Could this error be an issue with daemondo?

This really looks like a file permissions error. Unix domain sockets are files and have user permissions just like any other. It looks as though the OSX user attempting to access the database does not have file permissions to access the socket file. To confirm this I've done some tests on Ubuntu and psql to try to generate the same error (included below).

You need to check the permissions on the socket file and its directories /var and /var/pgsql_socket. Your Rails app (OSX user) must have execute (x) permissions on these directories (preferably grant everyone permissions) and the socket should have full permissions (wrx). You can use ls -lAd <file> to check these, and if any of them are a symlink you need to check the file or dir the link points to.

You can change the permissions on the dir for youself, but the socket is configured by postgres in postgresql.conf. This can be found in the same directory as pg_hba.conf (You'll have to figure out which one). Once you've set the permissions you will need to restart postgresql.

# postgresql.conf should contain...
unix_socket_directory = '/var/run/postgresql'       # dont worry if yours is different
#unix_socket_group = ''                             # default is fine here
#unix_socket_permissions = 0777                     # check this one and uncomment if necessary.

---

EDIT:

I've done a quick search on google which you may wish to look into to see if it is relavent. This might well result in any attempt to find your config file failing.

http://www.postgresqlformac.com/server/howto_edit_postgresql_confi.html

---

Error messages:

User not found in pg_hba.conf

psql: FATAL:  no pg_hba.conf entry for host "[local]", user "couling", database "main", SSL off

User failed password auth:

psql: FATAL:  password authentication failed for user "couling"

Missing unix socket file:

psql: could not connect to server: No such file or directory
	Is the server running locally and accepting
	connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?

Unix socket exists, but server not listening to it.

psql: could not connect to server: Connection refused
	Is the server running locally and accepting
	connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?

Bad file permissions on unix socket file:

psql: could not connect to server: Permission denied
	Is the server running locally and accepting
	connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?

My gut feeling is that this is (again) a mac/OSX-thing: the front end and the back end assume a different location for the unix-domain socket (which functions as a rendezvous point).

Checklist:

• Is postgres running: ps aux | grep postgres | grep -v grep should do the trick
• Where is the socket located: find / -name .s.PGSQL.5432 -ls (the socket used to be in /tmp; you could start looking there)
• even if you locate the (unix-domain) socket, the client could use a different location. (this happens if you mix distributions, or of you have a distribution installed someplace and have another (eg from source) installation elsewhere), with client and server using different rendez-vous addresses.

If postgres is running, and the socket actually exists, you could use:

• psql -h /the/directory/where/the/socket/was/found mydbname

(which attempts to connect to the unix-domain socket)

; you should now get the psql prompt: try \d and then \q to quit. You could also try:

• psql -h localhost mydbname.

(which attempts to connect to localhost (127.0.0.1)

If these attempts fail because of insufficient authorisation, you could alter pg_hba.conf (and SIGHUP or restart) In this case: also check the logs.

A similar question: https://stackoverflow.com/questions/8327680/cant-get-postgres-started/8328470#8328470

Note: If you can get to the psql prompt, the quick fix to this problem is just to change your config/database.yml, add:

host: localhost

or you could try adding:

host: /the/directory/where/the/socket/was/found

In my case, host: /tmp

Try uninstalling the pg gem (gem uninstall pg) then reinstalling -- if you use bundler, then bundle install, else gem install pg. Also, make sure path picks up the right version: Lion has a version of posgresql (prior versions didn't) and it may be in the path before your locally installed version (e.g. MacPorts, homebrew).

In my case: homebrew install of postgresql, updated postgresql, rails, etc. and then got this error. Uninstalling and reinstalling the pg gem did it for me.

The location of the socket file is baked into the gem at compile time. Thus you need to rebuild your pg gem.

gem pristine pg
# or
bundle exec gem pristine pg

This should resolve that particular issue.

If you're getting a similar error:

psql: could not connect to server: No such file or directory
Is the server running locally and accepting
connections on Unix domain socket "/tmp/.s.PGSQL.5432"?

This might do the trick (it did for me):

initdb /usr/local/var/postgres -E utf8

The directory specified should be different if you're not using OSX/Brew.

Note: This is not the exact error message seen above, but this thread is the first result for that error message.

what resolved this error for me was deleting a file called postmaster.pid in the postgres directory. please see my question/answer using the following link for step by step instructions. my issue was not related to file permissions:

https://stackoverflow.com/questions/13573204/psql-could-not-connect-to-server-no-such-file-or-directory-mac-os-x/13573207#13573207

the people answering this question dropped a lot of game though, thanks for that! i upvoted all i could

This is how I solved that error message, based partly on wildplasser's answer.

find / -name .s.PGSQL.5432 -ls 2> /dev/null
  => ... /tmp/.s.PGSQL.5432

So, there's my socket or whatever, but the client looks for it at:

/var/run/postgresql/.s.PGSQL.5432

So quite simply make a symbolic link to the /tmp/.s.PGSQL.5432:

sudo ln -s /tmp/.s.PGSQL.5432 /var/run/postgresql/.s.PGSQL.5432

Hope this helps to anyone. The seems kind of wrong, but hey, it works!

I started getting this after upgrading to a new postgres - I didn't realize I had hold data files.

First I tried to start the postgres server:

postgres -D /usr/local/var/postgres

which is how I saw this error

FATAL:  database files are incompatible with server
DETAIL:  The data directory was initialized by PostgreSQL version 9.0, which is not compatible with this version 9.3.5.

So then I found this answer on SO - related to an incompatibility error: https://serverfault.com/questions/342626/how-do-i-upgrade-postgresl-database-incompatibility-error

This is what fixed it

mv /usr/local/var/postgres /usr/local/var/postgres.old
initdb -D /usr/local/var/postgres

Just confirming I had a similar issue on PSQL and Django,

Looked like because my psql server was not shut down correctly and the postmaster.pid file was still present (should be deleted on proper shutdown automatically) in my postgres folder.

Deleted this and all good

I was getting this same error (it turns out it was an error with postmaster.pid. Here's how I got postgres up and running again (credit to Ricardo Burillo for the fix):

$ rm /usr/local/var/postgres/postmaster.pid 
$ pg_resetxlog -f /usr/local/var/postgres

I had similar problem when trying to use postgresql with rails. Updating my Gemfile to use new version of gem pg solve this problem for me. (gem pg version 0.16.0 works). In the Gemfile use:

gem 'pg', '0.16.0'

then run the following to update the gem

bundle install --without production
bundle update
bundle install

This happened to me today after my Macbook's battery died. I think this can be caused by improper shutdown. All you have to do in cases such as mine is delete postmaster.pid

Navigate to the folder

cd /usr/local/var/postgres

Check to see if postmaster.pid is present

ls

Remove postmaster.pid

rm postmaster.pid

> psql: could not connect to server: No such file or directory > Is the server running locally and accepting > connections on Unix domain socket "/tmp/.s.PGSQL.5432"?

I searching the solution for a while. So, this one fixed the issue for me as well (reinit db):

rm -r /usr/local/var/postgres  
initdb /usr/local/var/postgres -E utf8  
pg_ctl -D /usr/local/var/postgres -l logfile start

I use OS X 10.11.3 with brew.

I read many topics about this error and the solution to me was to simply restart the postgres with:

sudo service postgresql restart

Which is not mentioned here.

In my case none of previous solutions was good. Instead of using socket, you can use TCP host + port number in Rails config file. So in database.yml file just add two lines like here:

...
adapter: postgresql
encoding: unicode
pool: 5
host: localhost
port: 5432

This solved my problem :)

Before I used this fix:

sudo mkdir /var/run/postgresql
sudo ln -s /tmp/.s.PGSQL.5432 /var/run/postgresql/.s.PGSQL.5432

But after each reboot /tmp/.s.PGSQL.5432 was deleted and I had to repeat these commands. Solution works, but it is horrible, so better just modify Rails database config file :)

Got this error when I was setting up Posgtres with Django, I'm using Back Track and it comes with Postgres installed. I assume the settings are the issue. I fixed it by removing it completely then reinstalling like so.

sudo apt-get remove postgresql
sudo apt-get purge postgresql

Now run:

apt-get --purge remove postgresql\*

to remove everything PostgreSQL from your system. Just purging the postgres package isn't enough since it's just an empty meta-package.

Once all PostgreSQL packages have been removed, run:

rm -r /etc/postgresql/
rm -r /etc/postgresql-common/
rm -r /var/lib/postgresql/
userdel -r postgres
groupdel postgres

You should now be able to:

apt-get install postgresql

MacOSX here. I had the same problem after upgrading my postresql install from a pre-9.1 to 9.1.2 using homebrew. (By the way, remember to dump databases before your upgrade with pg_dump, pre-9.1 databases are incompatible.) Same problem, same error messages.

Uninstalling the pg gem did the trick for me. I actually had to do quite a bit of dancing to discover the issue. First I did a global gem uninstall, clearing the deck of all the old gems (there were a few). Then I removed pg from my Gemfile, rebundled, restored the pg reference, and rebounded once more.

After that, it worked like a charm.

Hello world :)
The best but strange way for me was to do next things.

1) Download postgres93.app or other version. Add this app into /Applications/ folder.

2) Add a row (command) into the file .bash_profile (which is in my home directory):

export PATH=/Applications/Postgres93.app/Contents/MacOS/bin/:$PATH

It's a PATH to psql from Postgres93.app. The row (command) runs every time console is started.

3) Launch Postgres93.app from /Applications/ folder. It starts a local server (port is "5432" and host is "localhost").

4) After all of this manipulations I was glad to run $ createuser -SRDP user_name and other commands and to see that it worked! Postgres93.app can be made to run every time your system starts.

5) Also if you wanna see your databases graphically you should install PG Commander.app. It's good way to see your postgres DB as pretty data-tables

Of, course, it's helpful only for local server. I will be glad if this instructions help others who has faced with this problem.

gem uninstall pg

On OS X with Homebrew:

gem install pg -- --with-pg-config=/usr/local/bin/pg_config

I had this problem plaguing me, and upon further investigation (running rake db:setup), I saw that rails was trying to connect to a previously used postgres instance - one which was stored in env variables as DATABASE_URL.

The fix: unset DATABASE_URL

source: https://stackoverflow.com/a/17420624/2577622

I tried most of the solutions to this problem but couldn't get any to work.

I ran lsof -P | grep ':5432' | awk '{print $2}' which showed the PID of the process running. However I couldn't kill it with kill -9 <pid>.

When I ran pkill postgresql the process finally stopped. Hope this helps.