Re: Multicolun index creation never completes on 9.0.1/solaris

On 1/26/11 12:13 PM, Tom Lane wrote:
> Josh Berkus <josh@agliodbs.com> writes:
>> Oh!  Actually, it only *did* 27 runs.  So it actually completed building
>> the index.  I'd expected trace_sort to give me some kind of completion
>> message; apologies for not checking all screen windows.
>
> Huh, there should be an "external sort ended" message, as well as some
> other ones once it finishes writing the initial runs.

Oh, fooled by log rotation:

LOG:  finished writing run 83 to tape 82: CPU 2600.37s/19449.04u sec
elapsed 22402.36 sec

LOG:  finished writing run 84 to tape 83: CPU 2630.31s/19674.93u sec
elapsed 22659.40 sec

LOG:  finished writing run 85 to tape 84: CPU 2658.46s/19905.44u sec
elapsed 22919.25 sec

LOG:  performsort starting: CPU 2686.35s/20121.46u sec elapsed 23164.61 sec

LOG:  finished writing run 86 to tape 85: CPU 2686.48s/20127.46u sec
elapsed 23170.74 sec

LOG:  finished writing final run 87 to tape 86: CPU 2688.70s/20220.72u
sec elapsed 23266.35 sec

LOG:  performsort done (except 87-way final merge): CPU
2691.55s/20224.35u sec elapsed 23277.44 sec

LOG:  external sort ended, 7991989 disk blocks used: CPU
3237.80s/27221.09u sec elapsed 31382.50 sec

So, a total of almost 9 hours.  Clearly there's something wrong here; if
I have time I'll try a recompile with GCC and see how it does.
Collation is 'C'.

--
                                  -- Josh Berkus
                                     PostgreSQL Experts Inc.
                                     http://www.pgexperts.com