Ed,
My own notes on how to get timestamped error logs refer to the file
"...src/include/utils/elog.h". There is a commented out "#define
ELOG_TIMESTAMPS" line in that header as well. Since version 6.4, I have
always uncommented THAT line and I get timestamps in my error logs... can't
tell you wh
Oh, one more thing. I've always made the edit BEFORE running ./configure.
Phil Culberson
-Original Message-
From: Culberson, Philip
Sent: Friday, December 17, 1999 11:42 AM
To: 'Ed Loehr'; [EMAIL PROTECTED]
Subject: RE: [GENERAL] How to get timestamped pgsql logging
I redirect all output from the
postmaster into a file with the following command:
exec postmaster -i -B 12000 -d 2 -o "-F -S 4096 -s" >&! server.log &
Phil Culberson
-Original Message-
From: Ed Loehr [mailto:[EMAIL PROTECTED]]
Sent: Friday, December 17, 1999 2:21 PM
To: Cu
>> postmaster -i -o "-F -S 4096 -s"
Remove the "-s" from the backend options.
Happy New Year
Phil Culberson
-Original Message-
From: Ed Loehr [mailto:[EMAIL PROTECTED]]
Sent: Sunday, January 02, 2000 4:18 PM
To: pg-gen
Subject: [GENERAL] How 2 shutoff verbose logging...
Could someo
In his very insightful post last week, Mike Mascari pointed out that, on
tables with heavy insert/updates, it was much faster to drop the index,
vacuum analyze, and then rebuild the index. Maybe in vacuum there is a
specific inefficiency in what Mike coined "defragment"ing indexes.
[Snip]
8. No
It seems to me that in this case Bruce would be better off to use a default
value and NOT "simulate" an outer join.
I suggest the following:
Instead of using a character abbreviation for the relation, use a number.
Since the list of categories is most likely going to remain small, you can
use an
JB,
The 20 seconds definitely sounds excessive. Have you done the following?
1) Run "vacuum analyze info"? If you have not, Postgres will not make use
of any indices.
2) Run an explain plan on your query to see what Postgres thinks it's going
to do?
In another post, Howie suggested more RAM
Maybe you didn't send the right snippet from your SQL, but your example
shows you building an index on "ref_ref_inspec", and the query you are
having a problem with has a WHERE clause referencing "ref_ref_article".
-Original Message-
From: Peter Haworth [mailto:[EMAIL PROTECTED]]
Sent: T
I ran a couple of tests against an Oracle 7.3.4 database.
I inserted a row into a table, then immediately typed exit. The insert was
committed automatically.
I also did an insert and then killed SQL*Plus with a SIGKILL from another
window. The transaction was NOT committed.
So, by default, SQ
Gabriel,
If you defined column "code" as a CHAR(n), you will need to have a "%"
wildcard at the end of your search string to allow for padding.
select code from codes where code like '(4)%' ;
Example:
business=> create table foo ( mytext char(10) );
CREATE
business=> insert into foo values ('4'
hing has changed between my version and 6.5.3.
Are you running your queries through psql? Also, is it possible that there
are leading characters in the data?
Phil Culberson
DAT Services
-Original Message-
From: Gabriel Fernandez [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, March 21, 2
It appears to function properly in 7.0 Beta 2:
philtst=# select version();
version
--
PostgreSQL 7.0.0 on sparc-sun-solaris2.7, compiled by gcc 2.95.2
(1 row)
philtst=# SEL
Since triggers ARE fired by the COPY command, another possibility is the use
of a temporary trigger to simulate the functionality of DEFAULT.
Just change the name of the sequence and the table in the example below.
Once you are done with your copy, simply DROP TRIGGER fake_sequence ON foo;
Cavea
You cannot use a "name" in a WHERE clause and aliases are specifically for
aliasing table names only. (An alias reference would look like "WHERE
alias.column_name > 10".) The documentation for "AS name" usage specifically
says a "name" cannot be used in a WHERE clause. I believe you will have to
This reminds me of a problem I once had.
I was trying to "make" the documentation, but zcat kept failing because the
Solaris version does not work the same as the GNU version. So I installed
the GNU zcat, ran configure again, but still make was failing because of
zcat...
I found that once confi
I think what you need to do is:
$ initlocation $P
^
$ createdb indybook -D $P
^
-Original Message-
From: Richard J Kuhns [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, May 16, 2000 8:26 AM
To: [EMAIL PROTECTED]
Subject: [GENERAL] Question about databases
I've worked with various versions of Oracle for several years and can share
some of my experiences with their "system catalog" implementation.
They use a fairly simple design in which a database instance consists of 1
.. n tablespaces (that can contain any type of database object) which in
turn c
17 matches
Mail list logo
