Bug #2835
closed
Rudder 2.4.0~beta3 fails to install Rudder server at rudder-reports stage with error message about "connections on Unix domain socket"
Added by Nicolas PERRON over 11 years ago.
Updated about 9 years ago.
Description
During installation, rudder-reports fails with these messages:
[...]
Setting postgresql as a boot service
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"?
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"?
/var/tmp/rpm-tmp.aVk9Hl: line 19: [: -ne : opérateur unaire attendu
/var/tmp/rpm-tmp.aVk9Hl: line 22: [: -ne : opérateur unaire attendu
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"?
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"?
psql: n'a pas pu se connecter au serveur : Connexion refusée
Le serveur est-il actif sur l'hôte « localhost » et accepte-t-il les connexions
TCP/IP sur le port 5432 ?
warning: %post(rudder-reports-1344966605:2.4.0.beta3-1.EL.6.noarch) scriptlet failed, exit status 2
[...]
It results in a rudder database which is missing and the Webapp can't be started:
[...]
Caused by:
org.apache.commons.dbcp.SQLNestedException: Cannot create PoolableConnectionFactory (FATAL: password authentication failed for user "rudder")
[...]
[...]
R: @@DistributePolicy@@result_success@@root-DP@@root-distributePolicy@@00@@reports@@None@@2012-08-22 12:15:56-04:00##root@#The OpenLDAP configuration file is OK (checked rootdn password)
psql: FATAL: password authentication failed for user "rudder"
-> Executing '/usr/bin/psql -q -c "ALTER USER rudder WITH PASSWORD '1195d7218ad6'"' ...(timeout=-678,owner=26,group=-1)
!! Finished command related to promiser "/usr/bin/psql -q -c "ALTER USER rudder WITH PASSWORD '1195d7218ad6'"" -- an error occurred (returned 1)
Q: "...r/bin/psql -q -": ERROR: role "rudder" does not exist
[...]
- Status changed from New to In progress
Trying to use rpm with extra verbose mode result in:
+ /etc/init.d/postgresql start
Démarrage du service postgresql : [ OK ]
+ echo 'Setting postgresql as a boot service'
Setting postgresql as a boot service
+ /sbin/chkconfig --add postgresql
+ /sbin/chkconfig postgresql on
+ dbname=rudder
+ usrname=rudder
++ su - postgres -c 'psql -t -c "select count(1) from pg_catalog.pg_database where datname = '\''rudder'\''"'
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"?
+ RES=
++ su - postgres -c 'psql -t -c "select count(1) from pg_user where usename = '\''rudder'\''"'
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"?
+ RES2=
+ '[' -ne 0 ']'
/var/tmp/rpm-tmp.Z20J9I: line 20: [: -ne : opérateur unaire attendu
+ '[' -ne 0 ']'
/var/tmp/rpm-tmp.Z20J9I: line 23: [: -ne : opérateur unaire attendu
+ su - postgres -c 'psql -q -c "CREATE USER rudder WITH PASSWORD '\''Normation'\''"'
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"?
+ su - postgres -c 'psql -q -c "CREATE DATABASE rudder WITH OWNER = rudder"'
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"?
+ echo localhost:5432:rudder:rudder:Normation
+ chmod 600 /root/.pgpass
+ psql -q -U rudder -h localhost -d rudder -f /opt/rudder/etc/postgresql/reportsSchema.sql
psql: n'a pas pu se connecter au serveur : Connexion refusée
Le serveur est-il actif sur l'hôte « localhost » et accepte-t-il les connexions
TCP/IP sur le port 5432 ?
D: install: waitpid(1552) rc 1552 status 200 secs 16.082
attention: %post(rudder-reports-1344966605:2.4.0.beta3-1.EL.6.noarch) scriptlet failed, exit status 2
If I try to reuse some of these commands:
[root@vagrant-centos-6 ~]# su - postgres -c 'psql -t -c "select count(1) from pg_catalog.pg_database where datname = '\''rudder'\''"'
0
[root@vagrant-centos-6 ~]# su - postgres -c 'psql -t -c "select count(1) from pg_user where usename = '\''rudder'\''"'
0
It works perfectly.... What is needed could be a loop to wait until postgresql is truely running.
I just saw your commit to change the postinst script in packaging, and am rather bothered by this line:
+while [ `ps ax | grep post | grep -v "grep" | wc -l` -lt 5 ]; do echo -n "."; done
This is very lazy matching! "grep post" will match hundreds of processes, and maybe occasionaly the one you want. If you know what process name you're looking for, you should at least use it here (ie /usr/bin/postgresql-master or whatever), but still this is ugly. If you need to test whether you can use the "psql" command, why not just try running a simple query with psql and checking the return code in a while loop until it's 0 ?
Jonathan CLARKE wrote:
I just saw your commit to change the postinst script in packaging, and am rather bothered by this line:
[...]
This is very lazy matching! "grep post" will match hundreds of processes, and maybe occasionaly the one you want. If you know what process name you're looking for, you should at least use it here (ie /usr/bin/postgresql-master or whatever), but still this is ugly. If you need to test whether you can use the "psql" command, why not just try running a simple query with psql and checking the return code in a while loop until it's 0 ?
You're right.
We should use a command like this one:
su postgres -c "psql -q --output /dev/null -c \"SELECT COUNT(*) FROM pg_catalog.pg_authid\""
I have to search how to use it il a while loop.
- Target version changed from 2.4.0~beta4 to 2.4.0~beta5
- Status changed from In progress to Pending technical review
- % Done changed from 0 to 100
Applied in changeset commit:a2be68d33c392babf67a10d2350c3ad560e8b7a2.
- Status changed from Pending technical review to Discussion
Nicolas PERRON wrote:
Applied in changeset commit:a2be68d33c392babf67a10d2350c3ad560e8b7a2.
This looks much better Nicolas, thank you.
To go one step further, we should ensure we don't get stuck in this loop forever. The current script will wait in the while loop until PostgreSQL is up... But what if it never comes up? This will block packaging completely. I suggest we fix a timeout, let's say 60 seconds, add a "sleep 1" to each iteration, and use a counter to measure how many times we've been through the loop and "break" from the loop if the timeout is reached.
- Target version changed from 2.4.0~beta5 to 2.4.0~beta4
All these commits (including the next one you're about to do :) ) should be included in beta4.
Last commit looks good - just needs an extra line at the end, in case the while loop succeeds, and prints "....", you need to add:
echo ""
Jonathan CLARKE wrote:
Last commit looks good - just needs an extra line at the end, in case the while loop succeeds, and prints "....", you need to add:
[...]
It should not be needed as I have use this:
echo -e "\nConnection to PostgreSQL has not been established before timeout"
Nicolas PERRON wrote:
Jonathan CLARKE wrote:
Last commit looks good - just needs an extra line at the end, in case the while loop succeeds, and prints "....", you need to add:
[...]
It should not be needed as I have use this:
[...]
Yes, but that's only displayed for a failure. What about a success?
Jonathan CLARKE wrote:
Nicolas PERRON wrote:
Jonathan CLARKE wrote:
Last commit looks good - just needs an extra line at the end, in case the while loop succeeds, and prints "....", you need to add:
[...]
It should not be needed as I have use this:
[...]
Yes, but that's only displayed for a failure. What about a success?
Huh... you're right. I'll correct this
- Status changed from Discussion to Released
Thanks, looks good to me.
- Status changed from Released to New
- % Done changed from 100 to 90
This bug has not been fixed
- Status changed from New to Pending technical review
Matthieu or Jon, could you review the last commit (3ee97fb9b3117b1a348bfa38f8afb64d0d3677a1), please ?
- Status changed from Pending technical review to Released
- % Done changed from 90 to 100
Nicolas PERRON wrote:
Matthieu or Jon, could you review the last commit (3ee97fb9b3117b1a348bfa38f8afb64d0d3677a1), please ?
OK for me!
- Project changed from Rudder to 34
- Category deleted (
11)
- Project changed from 34 to Rudder
- Category set to Packaging
Also available in: Atom
PDF
Updated by Nicolas PERRON 
Updated by 
Updated by