eninge-cleanup left db non-empty (was: [Users] Downgrading to 3.3.3 after 3.4.0 beta 2 testing)

[Yedidyah Bar David]

----- Original Message -----
> From: "Trey Dockendorf" <treydock at gmail.com>
> To: "Yedidyah Bar David" <didi at redhat.com>
> Cc: "users" <users at ovirt.org>, "infra" <infra at ovirt.org>
> Sent: Thursday, February 13, 2014 7:36:52 PM
> Subject: Re: eninge-cleanup left db non-empty (was: [Users] Downgrading to 3.3.3 after 3.4.0 beta 2 testing)
> 
> On Thu, Feb 13, 2014 at 1:55 AM, Yedidyah Bar David <didi at redhat.com> wrote:
> > ----- Original Message -----
> >> From: "Yedidyah Bar David" <didi at redhat.com>
> >> To: "Trey Dockendorf" <treydock at gmail.com>
> >> Cc: "users" <users at ovirt.org>, infra at ovirt.org
> >> Sent: Thursday, February 13, 2014 8:32:01 AM
> >> Subject: eninge-cleanup left db non-empty (was: [Users] Downgrading to
> >> 3.3.3 after 3.4.0 beta 2 testing)
> >>
> >> ----- Original Message -----
> >> > From: "Trey Dockendorf" <treydock at gmail.com>
> >> > To: "users" <users at ovirt.org>
> >> > Sent: Wednesday, February 12, 2014 8:19:25 PM
> >> > Subject: [Users] Downgrading to 3.3.3 after 3.4.0 beta 2 testing
> >> > $ engine-cleanup
> >> > $ engine-backup --mode=restore --scope=all
> >> > --file=engine-20140211-1457.tar.bz2 --log=engine-backup.log
> >> > Restoring...
> >> > FATAL: Database is not empty
> >> >
> >> > Attached are the two engine-cleanup logs from both attempts.  The
> >> > first (from 20140211) was answering "Yes" only to "remove Engine DB
> >> > content".  The second (from 20140212, today) was "Yes" to "remove all
> >> > components".
> >>
> >> Apparently engine-cleanup does not clean up everything. We tried to
> >> make it do that, and I am pretty certain it used to at some point...
> >>
> >> 1. You might want to open a bug about this. As you already posted,
> >> manually dropping and creating the database still works...
> >> 2. Adding infra@ - I think we should add a jenkins job to verify
> >> that engine-cleanup cleans up at least the database, perhaps other
> >> things. I am pretty certain it should be so for 3.3, didn't check
> >> 3.4 yet.
> >
> > Well, I now tried that with 3.4.0-beta2 and did not manage to reproduce -
> > database was empty after engine-cleanup.
> >
> > If you manage to reproduce, please post the output of:
> > pg_dump engine | grep -i ^create
> > (as postgres, or passing credentials as needed).
> > It should only output one line:
> > CREATE EXTENSION IF NOT EXISTS plpgsql WITH SCHEMA pg_catalog;
> >
> > If it outputs anything else it's probably a bug.
> >
> > Thanks, FYI,
> > --
> > Didi
> 
> With 3.3.3 I got this after engine-cleanup.
> 
> (/root/pgpass was created by me mimicking what's created by engine-backup).
> 
> $ PGPASSFILE=/root/pgpass pg_dump -U engine -h localhost -p 5432
> engine | grep -vi '^create extension' | grep -i '^create'
> CREATE PROCEDURAL LANGUAGE plpgsql;

OK, makes (some?) sense...
Eli - How come it does not always appear in pg_dump's outout? Is it because
it's not always created, or not always output, or something else?

> 
> The grep statements above mimic what was executed by the engine-backup
> which resulted in a message "FATAL: Database is not empty".
> 
> I will try and reproduce with 3.4.0-beta2.
> 
> Should I file a bug for this issue against 3.3.x?

Please do!

Thanks for the report,
-- 
Didi