Re: BUG #13813: pg_upgrade fails to run pg_ctl - Mailing list pgsql-bugs

From Michael Paquier
Subject Re: BUG #13813: pg_upgrade fails to run pg_ctl
Date
Msg-id CAB7nPqTU19HG9wUup1qrTm3Vxnw9uMa48hchFBE1Xs=V0NpKEA@mail.gmail.com
Whole thread Raw
In response to BUG #13813: pg_upgrade fails to run pg_ctl  (romandashkov@gmail.com)
List pgsql-bugs
On Sun, Dec 13, 2015 at 1:37 PM,  <romandashkov@gmail.com> wrote:
> I ran pg_upgrade as a test, trying to move data files from 9.4.5 to 9.4.5
> (same version in two separate directories). Both old and new servers were
> off. Command issued:
>
> pg_upgrade --username postgres --no-paswword -old-datadir... -new-datadir...
> -old-bindir... -new-bindir...
>
> It failed with this a message like this:
>
> could not start old postmaster... with this command...
> "path\to\data-directory/pg_ctl" -option 1 -option 2...
>
> I noticed it is using a forward slash as a path to pg_ctl within the data
> directory. It should be using backward slash in windows.

Both are recognized on Windows so that's not really an issue to mix them.

> I'm sorry I was not able to record the error message exactly before it was
> lost (and the logs too).

It would be good to get more information, or it will be difficult to
diagnose a problem, something that got fixed lately for pg_upgrade on
Windows is for example the detection of copy failures:
commit: 025106e314627f5aacdf72fca52ba0ed985d696e
author: Bruce Momjian <bruce@momjian.us>
date: Sat, 14 Nov 2015 11:47:12 -0500
pg_upgrade:  properly detect file copy failure on Windows

Previously, file copy failures were ignored on Windows due to an
incorrect return value check.

Perhaps you got bitten by that... Or not.

Regards,
--
Michael

pgsql-bugs by date:

Previous
From: Michael Paquier
Date:
Subject: Re: Error postgres9.5
Next
From: Amit Kapila
Date:
Subject: Re: PQexec() hangs on OOM