Re: Abnormal termination of PostgreSQL

Tom Lane wrote:
> If you have the time and interest, it might be worth repeating the
> whole setup sequence starting from initdb. I'm speculating that
> installing PostGIS might have somehow left corruption in server memory
> that later manifested as the visible problem.
>

Tom,

I'm not sure if I did this right, but a rebuild of that data cluster
worked fine. Here's the steps I took (please correct me if I messed
something up here):

I stopped the server, then just deleted /usr/local/pgsql/data, followed by:

# su - postgres
$ /usr/local/pgsql/bin/initdb -D /usr/local/pgsql/data

Then I updated pg_hba.conf to the correct access settings, and started
the postmaster service again.

Following that I logged into psql and set a password for postgres and
created a lower-privileged user account (bthoen) to use with the
database I will be building (vfm).

Then I created the database and "spatialized" it.
$ createdb -O bthoen vfm
$ createlang plpgsql vfm
$ psql -f /usr/local/pgsql/share/lwpostgis.sql -d vfm

Finally I created the corn0715 table and populated it, ran my update
script and indexed the table. Finally I did a 'vacuum full' on it to
pack it back down.

The table seems fine this time, and there was no segfaulting nonsense at
any time. But the server log was a lot more active. Whe I loaded this
data the first time and had that corruption prroblem, there was hardly
anything posted to the log. However, this time I noticed when I ran the
postgis script that there was some weirdness written to the log that
doesn't look right . Does this look like geos or postgis is a bit broken
to you?

NOTICE: type "histogram2d" is not yet defined
DETAIL: Creating a shell type definition.
NOTICE: return type histogram2d is only a shell
NOTICE: argument type histogram2d is only a shell
NOTICE: argument type histogram2d is only a shell
NOTICE: type "spheroid" is not yet defined
DETAIL: Creating a shell type definition.
NOTICE: return type spheroid is only a shell
NOTICE: argument type spheroid is only a shell
NOTICE: argument type spheroid is only a shell
NOTICE: type "geometry" is not yet defined
DETAIL: Creating a shell type definition.
NOTICE: return type geometry is only a shell
NOTICE: argument type geometry is only a shell
NOTICE: argument type geometry is only a shell
NOTICE: return type geometry is only a shell
NOTICE: return type geometry is only a shell
NOTICE: argument type geometry is only a shell
NOTICE: argument type geometry is only a shell
NOTICE: type "box3d" is not yet definedDETAIL: Creating a shell type
definition.
NOTICE: argument type box3d is only a shell
NOTICE: return type box3d is only a shell
NOTICE: argument type box3d is only a shell
NOTICE: type "chip" is not yet defined
DETAIL: Creating a shell type definition.
NOTICE: return type chip is only a shell
NOTICE: argument type chip is only a shell
NOTICE: argument type chip is only a shell
NOTICE: type "box2d" is not yet defined
DETAIL: Creating a shell type definition.
NOTICE: return type box2d is only a shell
NOTICE: argument type box2d is only a shell
NOTICE: argument type box2d is only a shell
NOTICE: CREATE TABLE / PRIMARY KEY will create implicit index
"spatial_ref_sys_pkey" for table "spatial_ref_sys"
NOTICE: CREATE TABLE / PRIMARY KEY will create implicit index
"geometry_columns_pk" for table "geometry_columns"
ERROR: table "corn0715" does not exist
STATEMENT: drop table corn0715 ;
LOG: checkpoints are occurring too frequently (25 seconds apart)
HINT: Consider increasing the configuration parameter
"checkpoint_segments".
LOG: checkpoints are occurring too frequently (25 seconds apart)
HINT: Consider increasing the configuration parameter
"checkpoint_segments".
. . .

(lots of checkpoints static with the COPY loading, but otherwise things
looked normal from here on.)

So I think postgres is all right, but I'm not so sure about postgis. I
did have problems trying to compile GEOS (and GDAL) so maybe that's
where the trouble lies.

Any thoughts?

- Bill Thoen