Re: [HACKERS] Problems doing a copy to a table - Mailing list pgsql-hackers

From jwieck@debis.com (Jan Wieck)
Subject Re: [HACKERS] Problems doing a copy to a table
Date
Msg-id m107PsU-000EBPC@orion.SAPserv.Hamburg.dsh.de
Whole thread Raw
In response to Problems doing a copy to a table  ("Chris Williams" <chris_d_williams@itd.sterling.com>)
List pgsql-hackers
Chris Williams wrote:

>
> I am creating a text file to do file dumps into postgres using the copy command.  The copy command
> executes fine but when I do a count on the table, I get zero records.  But when I do a vacuum
> command on that table, here is what is says
>
> sdbm=> vacuum verbose D_TERM_APPROACH;
> NOTICE:  --Relation d_term_approach--
> NOTICE:  Pages 490: Changed 0, Reapped 490, Empty 0, New 0; Tup 0: Vac 11530, Crash 11530, UnUsed 0,
> MinLen 0, MaxLen 0; Re-using: Free/Avail. Space 3964040/0; EndEmpty/Avail. Pages 490/0. Elapsed 0/1
> sec.
> NOTICE:  Rel d_term_approach: Pages: 490 --> 0.
>
> I know the 11530 is the correct number of tuples that should have been added to the table.  Any
> ideas on what may be happening or how I can debug this problem.  I am using version 6.4 on Solaris
> 2.5.1.

    The  notice  about  'Crash  11530' tells that the copy didn't
    execute as fine as you thought. Don't know  what  causes  the
    crash  though as long as I can't see some part of what you're
    trying to copy in.

    But as a matter of fact,  the  backend  copying  in  must  do
    something  very  bad  at  the  end  of  it's  COPY.  Thus the
    transaction doesn't commit correctly and vacuum  just  whipes
    out  the data which came from a crashed transaction. Whithout
    a correct committed transaction, your whole approach is a big
    noop.

    Please  try  to  reproduce the error with a smaller amount of
    (maybe anonymized data) and give a  reproducable  example  of
    what  you're  in detail do (table schema, test-data and exact
    copy command).


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#======================================== jwieck@debis.com (Jan Wieck) #

pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: [GENERAL] big bad join problems
Next
From: Goran Thyni
Date:
Subject: New patch (was: tough bug)