Re: pg_upgrade from 9.1 to 9.2 takes a really long time (compared to previous versions)?

On Wed, Nov 7, 2012 at 01:09:19PM -0800, Lonni J Friedman wrote:
> On Wed, Nov 7, 2012 at 1:07 PM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> > On Wed, Nov 7, 2012 at 01:06:19PM -0800, Lonni J Friedman wrote:
> >> On Wed, Nov 7, 2012 at 12:59 PM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> >> > On Wed, Nov 7, 2012 at 10:44:12AM -0800, Lonni J Friedman wrote:
> >> >> On Wed, Nov 7, 2012 at 10:42 AM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> >> >> > On Fri, Oct 19, 2012 at 02:44:05PM -0700, Lonni J Friedman wrote:
> >> >> >> Greetings,
> >> >> >> I'm in the process of planning for a production upgrade from 9.1.6 to
> >> >> >> 9.2.x (all Linux-x86-64). In my staging environment (which has the
> >> >> >> same versions), I kicked off pg_upgrade about 5 hours ago, and its
> >> >> >> still not done. It is making progress, so I don't think anything has
> >> >> >> gone wrong, beyond it taking much longer than anticipated.
> >> >> >>
> >> >> >> When I used pg_upgrade to go from 9.0.x to 9.1.x, it finished in just
> >> >> >> under an hour. There is admittedly about three times as much data (in
> >> >> >> terms of disk usage) now than when I upgraded from 9.0.x. Would that
> >> >> >> explain the increased time needed to do the upgrade? Or is there
> >> >> >> something about the upgrade to 9.2.x that requires a lot more time?
> >> >> >>
> >> >> >> I'm trying to understand if what I'm seeing is expected, normal
> >> >> >> behavior, or if something might not be right.
> >> >> >
> >> >> > Odd. How many object/tables do you have? I have just patched 9.2 to
> >> >> > improve upgrades for clusters with many objects.
> >> >>
> >> >> about 5000 tables spread across 5 databases.
> >> >
> >> > That should not take very long. Are you using link mode?
> >>
> >> Nope. The command that I used was:
> >> pg_upgrade -b /usr/pgsql-9.1/bin -B /usr/pgsql-9.2/bin -d
> >> /var/lib/pgsql/9.1/data -D /var/lib/pgsql/9.2/data
> >
> > Well, it must then copy all the data from old to new cluster --- that
> > could take a while.
>
> I guess so, but its hard to tell. It was sitting for quite a while
> with no output.

It should spin through the file names as they are copied. What was the
last output line before it hung?

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +