Bug 1217317

This is also issue when installing Satellite wihh embedded PostgreSQL.

+++ This bug was initially created as a clone of Bug #1182044 +++

Description of problem:
Rarely when you are doing `./install.pl --upgrade --managed-db` it might fail.


Version-Release number of selected component (if applicable):
Satellite-5.7.0-RHEL6-re20150108.2-x86_64


How reproducible:
rarely (seen once and done about 10 runs of the command)


Steps to Reproduce:
1. # /mnt/tmp/install.pl --upgrade --managed-db


Actual results:
# /mnt/tmp/install.pl --upgrade --managed-db
* Starting Red Hat Satellite installer.
* Performing pre-install checks.
* Pre-install checks complete.  Beginning installation.
* RHN Registration.
** Registration: System is already registered with RHN.  Not re-registering.
* Upgrading managed DB packages.
*** Upgrading embedded database.
There was a problem during embedded database upgrade.
  Exit value: 2.
Please examine /var/log/rhn/rhn_installation.log for more information.
# tail /var/log/rhn/rhn_installation.log
Running Transaction
  Installing : postgresql92-postgresql-upgrade-9.2.8-2.el6.x86_64           1/1 
  Verifying  : postgresql92-postgresql-upgrade-9.2.8-2.el6.x86_64           1/1 

Installed:
  postgresql92-postgresql-upgrade.x86_64 0:9.2.8-2.el6                          

Complete!
Starting postgresql service: [  OK  ]
psql: FATAL:  the database system is starting up


When I have re-ran /usr/bin/spacewalk-upgrade-postgresql manually, upgrade passed.


Expected results:
Should work


Additional info:
This is a code-path:

### /mnt/tmp/install.pl:
[...]
if ($opts{"upgrade"} and $opts{"managed-db"}) {
  install_satellite_repo(\%opts);
  print loc("* Upgrading managed DB packages.\n");
  install_embedded_postgresql(\%opts);
  Spacewalk::Setup::postgresql_upgrade_embedded_db();
  print loc("* Managed DB instance was successfully upgraded.\n");
  exit 0;
}
[...]

### /mnt/tmp/install/lib/Spacewalk/Setup.pm:
[...]
sub postgresql_upgrade_embedded_db {
    print loc("*** Upgrading embedded database.\n");
    system_or_exit(['/usr/bin/spacewalk-upgrade-postgresql', ''], 1, "There was a problem during embedded database upgrade.\n");
    return 0;
}
[...]

### /usr/bin/spacewalk-upgrade-postgresql:
[...]
service postgresql start
DB_TIMEZONE=$(echo "show timezone;" | psql -U $DB_USER -d $DB_NAME -h localhost | sed -n 3p | sed 's/ //g')
LOG_TIMEZONE=$(echo "show log_timezone;" | psql -U $DB_USER -d $DB_NAME -h localhost | sed -n 3p | sed 's/ //g')
service postgresql stop
[...]

So looks like we are trying to stop PostgreSQL 8.4 right after we have started and we are not checking result of the commands betweet that start and stop so its results can be empty or something.

--- Additional comment from Jan Hutař on 2015-02-04 00:48:23 EST ---

Reported bug 1188942, which might be relevant.

--- Additional comment from Jan Hutař on 2015-04-30 02:00:54 EDT ---

Pavel Raiskup mentioned PostgreSQL initscript configuration option which can achieve that. Problem is that it appears in rh-postgresql94 from RHSCL-2.0: bug 1188942 comment #3