RE: Database broken after using pgadmin 'backup' on OSX

Sadly, I don’t have a Mac at the moment so I can’t investigate the internals of “brew switch”; that does tell me something though – if “brew switch” fixes the problem then figuring out what it’s changing when it switches may help trace the actual issue.

In varying orders of possibilities:

1. Does “brew switch” have a “let’s be RIDICULOUSLY verbose” command line switch or environment variable?

2. I think OS X/Darwin have “strace” but I know that the BSD subsystems have the equivalent if that’s a Linux-ism?

3. Does OS X have Dtrace?

4. I wonder if it would be possible to see a snapshot of the underlying /usr/local (probably) before and after a “brew switch” (zfs snapshtos but I don’t think OS X is usually sitting on ZFS but if you have a full backup of /usr e.g. by OS X’s backup manager or you do daily backups [!] it may be possible?)

Given that “brew switch” does fix it, as you say, if you can figure out what it’s changing (I suspect it’s symlinking .dlyibs and .sos and other libraries) and then find out what other thing is changing that.

DSL

From: Conor McNally [mailto:theconor(at)gmail(dot)com]
Sent: Thursday, 29 June 2017 6:00 PM
To: pgadmin-support(at)lists(dot)postgresql(dot)org
Subject: Re: Database broken after using pgadmin 'backup' on OSX

Hi,

Yes I do have two versions of plpgsql.so but it wasn't a problem before.

>locate plpgsql.so
/usr/local/Cellar/postgresql/9.5.3/lib/postgresql/plpgsql.so
/usr/local/Cellar/postgresql/9.6.2/lib/postgresql/plpgsql.so

Previously I performed backup using the earlier version of pgAdmin 4 and before that I used pgAdmin 3.

All terminals are affected immediately after running the backup tool: before and after reboot. The "brew switch" command is the only thing I've found that fixes it.

I don't really know how brew works though. Do you think there something I can do to fix my set-up?

Cheers,

Conor

On Thu, Jun 29, 2017 at 10:32 PM, David Lloyd <lloy0076(at)adam(dot)com(dot)au <mailto:lloy0076(at)adam(dot)com(dot)au> > wrote:

By no means an expert but I see:

* An environment installed using 'brew'
* Some virt_env
* A mish-mash of libraries from different lib directories in the error

I know this sounds ridiculous, but I wonder if the OP performed the task
until such time they got the error. And then rebooted (or at least
logged out and logged in - i.e. clear their library paths, library cache
and so forth which is best guaranteed by a reboot but at least logging
out and maybe starting a new terminal).

I'd hedge a bet there's more than one `plgpsql.so` on the system and
they're subtly incompatible with each other.

DSL

> -----Original Message-----
> From: Dave Page [mailto:dpage(at)pgadmin(dot)org <mailto:dpage(at)pgadmin(dot)org> ]
> Sent: Thursday, 29 June 2017 5:20 PM
> To: Conor McNally <theconor(at)gmail(dot)com <mailto:theconor(at)gmail(dot)com> >
> Cc: pgadmin-support(at)lists(dot)postgresql(dot)org <mailto:pgadmin-support(at)lists(dot)postgresql(dot)org>
> Subject: Re: Database broken after using pgadmin 'backup' on OSX
>
>
>
> > On 29 Jun 2017, at 21:06, Conor McNally <theconor(at)gmail(dot)com <mailto:theconor(at)gmail(dot)com> >
> wrote:
> >
> > Hi,
> >
> > Can anyone help with the following problem?
> >
> > Since upgrading to pgAdmin 4 v1.5 on macOS Sierra (v10.12.5), it
seems
> my database is 'broken' every time I use the pgAdmin 'Backup' tool.
> >
> > Here's what happens:
> > * macOS Sierra v10.12.5
> > * postgres 9.6.2 locally installed with brew
> > * pgAdmin 4 v1.5 installed directly from website
> > * database is running fine
> > * use the pgAdmin Tools / Backup... utility to take a full back-up
> > * database is no longer working
> >
> > By 'no longer working' it means that the postrges libraries have
been
> messed up and I get the following error when I try to connect via
> pyscopg2 (python v3.5.2_3, pyscopg2 v2.6.2):
> >
> > File "/Users/conor/virt_env/conor/lib/python3.5/site-
> packages/psycopg2/extras.py", line 288, in execute
> > return super(NamedTupleCursor, self).execute(query, vars)
> > psycopg2.InternalError: could not load library
> "/usr/local/lib/postgresql/plpgsql.so":
> dlopen(/usr/local/lib/postgresql/plpgsql.so, 10): Symbol not found:
> _DatumIsReadWriteExpandedObject
> > Referenced from: /usr/local/lib/postgresql/plpgsql.so
> > Expected in: /usr/local/opt/postgresql/bin/postgres
> > in /usr/local/lib/postgresql/plpgsql.so
> >
> >
> > I find the only way to fix the problem is to run "brew switch
postgresql
> 9.6.2". I don't really know what that command does but the database
> connections are OK after I run it. Also, the database backup file has
been
> created successfully as far as I can tell: I have not yet attempted to
> restore it. One other thing is I don't get a confirmation message in
the
> pgAdmin console the way I used to.
> >
> > Can anybody help with the above problem? It has only started
> happening since I upgraded to pgAdmin 4 v1.5. Before that (with
> pgAdmin 4 v1.4) everything was fine.
>
> pgAdmin uses it's own private virtual environment (including libpq and
> friends). I honestly don't see how it could affect the system Python,
let
> alone a virtual envy from a completely different version.