I am aware that this or similar has been asked in other places. None of the other answers have worked for me.
Machine is a Macbook Pro 2014 / OS X Yosemite 10.10 with Postgres 9.3.5 installed via Homebrew.
I can only connect to my local postgres instance if I do NOT specify a host. This works:
psql -Umyuser -d mydb
sql (9.4.4)
Type "help" for help.
myuser=#
-h localhost does not:
psql -Umyuser -d mydb -h localhost
psql: server closed the connection unexpectedly
This probably means the server terminated abnormally
before or while processing the request.
It pauses for a long time before I get this message.
I am trying to set up a third party tool that seems to want to specify localhost, so it is not an option just to leave that off.
If I understand what I've read correctly elsewhere, postgres behaves differently if you're using TCP/IP, and specifying -h localhost forces that. Accordingly I have added the following in pg_hba.conf:
local all all trust
# IPv4 local connections:
host all all 127.0.0.1/32 trust
host all all localhost trust
# IPv6 local connections:
host all all ::1/128 trust
(127.0.0.1 AND localhost seems redundant but desperation drives superstition) ... and in postgresql.conf:
listen_addresses = 'localhost'
port = 5432
netstat confirms that Postgres is listening to that port:
☁ ~ netstat -an | grep 5432
tcp4 0 0 127.0.0.1.5432 *.* LISTEN
tcp6 0 0 ::1.5432 *.* LISTEN
d8d7c0e31d5add1d stream 0 0 d8d7c0e326b18a1d 0 0 0 /tmp/.s.PGSQL.5432
All permissions are set on the socket file:
~ ls -l /tmp/.s.PGSQL.5432
srwxrwxrwx 1 fritz wheel 0 12 Jul 13:27 /tmp/.s.PGSQL.5432
/etc/hosts has an entry for localhost:
127.0.0.1 localhost
::1 localhost
psql and pg_ctl are pointing to the same installations:
~ which psql
/usr/local/bin/psql
~ ll /usr/local/bin/psql
/usr/local/bin/psql -> ../Cellar/postgresql/9.4.4/bin/psql
~ which pg_ctl
/usr/local/bin/pg_ctl
~ ll /usr/local/bin/pg_ctl
/usr/local/bin/pg_ctl -> ../Cellar/postgresql/9.4.4/bin/pg_ctl
There's nothing named psql elsewhere on my system that seems like it could be hijacking this process:
~ sudo find / -name psql
/Applications/pgAdmin3.app/Contents/SharedSupport/psql
/Users/fritz/devtools/phantomjs/src/qt/config.tests/unix/psql
/Users/fritz/Qt/5.3/Src/qtbase/config.tests/unix/psql
/Users/fritz/Qt/5.3/Src/qtbase/src/plugins/sqldrivers/psql
/Users/fritz/Qt/5.3/Src/qtbase/src/sql/drivers/psql
/usr/local/bin/psql
/usr/local/Cellar/postgresql/9.4.4/bin/psql
UPDATE: telnet connects to localhost 5432 for a while and then is disconnected:
~ telnet localhost 5432
Trying ::1...
Connected to localhost.
Escape character is '^]'.
Connection closed by foreign host.
UPDATE 2: I used wireshark to spy on tcp localhost port 5432. Trouble is, I don't know what I should see and whether/how the results are wrong. Hoping someone out there can make something of this capture file.
2 Answers 2
Thanks to the suggestion on pgsql-general that I check TCP/IP with netcat I discovered that the first few bytes sent to localhost were being swallowed by a process that was intended to keep an audit trail of Internet access. Uninstalled and everything works as I expect.
There seems to be a special case in the PostgreSQL client libraries for localhost and 127.0.0.1. Instead of trying to hit the TCP port, it looks for the UNIX domain socket instead.
I was able to work around this by specifying localhost. as the hostname. (I was also able to get the IPv6 ::1 address to work.)
-h 127.0.0.1. The entry mentioned in/etc/hostsis probably not the only IP assigned tolocalhost.telnet 127.0.0.1 5432in the hope of getting a more meaningful error message.psqldoesn't just exit silently like that. If the connection is closed unexpectedly by the other end it'll print a message when you try the next query, not just exit. Similarly on authentication failures. I have a hard time guessing what might be going on here. In your position I'd be attaching a debugger, and/or doing network traffic capture, to figure out what's going on.