Re: Possible bug with shared memory buffers

Tom Lane wrote:
>
> Mark Rae <m(dot)rae(at)inpharmatica(dot)co(dot)uk> writes:
> > Normally the backend process would 'swap in' all 512M of shared
> > memory when loading, but occasionally after dropping the previous
> > database, the new backend would only seem to be able to use a
> > small amount of the shared memory.
>
> This is so vague that I'm not even sure what you're complaining about.
> What do you mean by "only seem to be able to use a small amount of the
> shared memory"?
>

I was referring to the memory usage as reported by top,ps etc.

Normally as the data is loaded, the RSS,SIZE,SHARE values for the
backend would grow towards ~512M.
However when the problem occurs the memory usage would not grow as normal
but get 'stuck' at some lower value (about 10M or so).

In 'Case 2' which I described, the process size was reporting
the use of all 512M, but it still slowed down in the same way.

i.e. the speed of loading was as expected, until the database
grew above a certain size, at which point the performance dropped
drastically. And at that point the whole database (including
indexes) certainly couldn't have been more than about 100M in size.

Also for information, the machine is dedicated machine, with nothing
else running except the loading process which is about 5M. So it
couldn't have been anything else hogging the memory.

-Mark

--
Mark Rae Tel: +44(0)20 7074 4648
Inpharmatica Fax: +44(0)20 7074 4700
m(dot)rae(at)inpharmatica(dot)co(dot)uk http://www.inpharmatica.co.uk/