PostgreSQL has to run before upgrading legacy installation

[SebastienGardoll]

Describe the bug

Legacy installation makes esgf node starts at boot time, with the command esg-node.
So as to rely only on the start/stop recipes of esgf-ansible, we have to disable this feature.
But disable this feature, reboot and than upgrade the legacy installation makes the upgrade to fail with this message:

TASK [base : Create db super user] ******************************************************************************
An exception occurred during task execution. To see the full traceback, use -vvv. The error was: socket Unix « /tmp/.s.PGSQL.5432 » ?
[WARNING]: Module remote_tmp /var/lib/pgsql/.ansible/tmp did not exist and was created with a mode of 0700,
this may cause issues when running as another user. To avoid this, create the remote_tmp dir with the correct
permissions manually
fatal: [vesgint-data.ipsl.upmc.fr]: FAILED! => {
"changed": false
}

MSG:

unable to connect to database: n'a pas pu se connecter au serveur : Aucun fichier ou dossier de ce type
Le serveur est-il actif localement et accepte-t-il les connexions sur la
socket Unix « /tmp/.s.PGSQL.5432 » ?

An exception occurred during task execution. To see the full traceback, use -vvv. The error was: socket Unix « /tmp/.s.PGSQL.5432 » ?
fatal: [vesgint-idx.ipsl.upmc.fr]: FAILED! => {
"changed": false
}

MSG:

unable to connect to database: n'a pas pu se connecter au serveur : Aucun fichier ou dossier de ce type
Le serveur est-il actif localement et accepte-t-il les connexions sur la
socket Unix « /tmp/.s.PGSQL.5432 » ?

To Reproduce
Steps to reproduce the behavior:

1. Disable esg-node start at boot time:

chmod -x /etc/init.d/zesgf
chmod -x /etc/init.d/esgf-httpd

2. reboot the machine

3. Upgrade a legacy installation

Expected behavior

esgf-ansible starts postgreSQL but it must start postgreSQL with the good configuration.

ESGF Node (please complete the following information):

• Managed OS: CentOS 6
• Host OS: CentOS 7
• esgf-ansible Version: v4.0.2
• Node type: a set of idp, index and data nodes

Additional context
legacy installation

[sashakames]

While not necessarily the cause of this issue, starting esg-node at boot time, as I see done here with creating the daemon zesgf has never to my knowledge been considered a standard or recommended practice. It may have worked out fine for your site (IPSL) in the past, but must have been done manually.

[SebastienGardoll]

esgf-ansible starts postgresql during the upgrading process and right after it tries to connect to it. I wonder if esgf-ansible is just too fast: I saw that rerun the installation recipe of esgf-ansible doesn't trigger the error. Only one time, I saw that the esgf-ansible succeeded to connect to the postgresql of the data node but not to the postgresql of the index node. Adding a pause after starting postgresql may solve this problem.

[SebastienGardoll]

FYI, I still have this issue when upgrading from 4.0.3 to 4.0.4 .