Re: Re: Heaps of read() syscalls by the postmaster

Hi,

Chris:
> Matthias Urlichs wrote:
>
> > You're right. It's the database's test/pg_attribute file,
> > which is a whopping 41 MBytes.
>
> I suppose the obvious question is whether you copy the database to a new
> database, if the new database's pg_attribute is 41MB.

? I don't understand.

The database was created by a simple initdb/createdb. The test user was
created and given access, and the benchmark was started. The person
watching the benchmark subsequently fell asleep. ;-)

The reason for the huge size of this file might be the fact that the
full benchmark first creates a whole damn lot of tables, which it then
deletes. Apparently this process results in a rather suboptimal
pg_attribute/pg_index file. It also leaks memory (about 2k per table)
in the backend.

I'll restart the whole thing later today. We'll see if the problem comes
back. Hopefully not.

--
Matthias Urlichs | noris network GmbH | smurf(at)noris(dot)de | ICQ: 20193661
The quote was selected randomly. Really. | http://smurf.noris.de/
--
Guitar players had their licks.