Peer authentication failed for user "postgres" on initial chef run (reboot + 2nd run works fine)

[nlochschmidt]

I am trying to setup a fairly simple Vagrantfile that boots a VM with postgresql and creates a couple of users and databases using hw-cookbooks/database. In the end I need access to the database from the host machine, so I change the pg_hba.conf by setting a couple of attributes in the Vagrantfile:

chef.json = {
  "postgresql" => {
    "version" => "9.3",
    "password" => {
      "postgres" => "postgres",
    },
    "config" => {
      "listen_addresses" => "*"
    },
    "pg_hba" => [
      { "type" => 'host', "db" => 'all', "user" => 'all', "addr" => '10.0.0.0/16', "method" => 'md5' },
      { "type" => 'local', "db" => 'all', "user" => 'all', "addr" => '', "method" => 'trust' }
    ]
  }
}

Now, when I boot the VM for the first time I get FATAL: Peer authentication failed for user "postgres". When I reboot the VM and run chef again everything works as it should, but I destroy and reprovision the VM a couple of times per month and it's just cumbersome.

I also see this in my log:

INFO: template[/etc/postgresql/9.3/main/pg_hba.conf] not queuing delayed action restart on service[postgresql] (delayed), as it's already been queued

So my guess is this has to do with the service notification having been changed from :immediately to :delayed and therefore when I try to create the users and databases during the initial convergence run the default pg_hba.conf is still in effect.

Am I doing something wrong? Help would be appreciated.

[chr4]

I'm having a similar issue - Creating databases in the same Chef run as configuring postgresql seems to me like a regular usecase.
Was there a particular reason the action was changed to :delayed?

+1

[nlochschmidt]

I experienced this on 3.4.14 but can not reproduce this on 3.4.20. So it seems to be fixed.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.