Commit log issue with load generated by pgbench

From: Gavin Sherry <swm(at)linuxworld(dot)com(dot)au>
To: pgsql-hackers-win32(at)postgresql(dot)org
Subject: Commit log issue with load generated by pgbench
Date: 2004-09-01 06:27:00
Message-ID: Pine.LNX.4.58.0409011622001.31854@linuxworld.com.au
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers-win32

Hi all,

I didn't see this mentioned in the archives and I've recreated it a few
times (could it be a configuration issue?).

I ran up the win32 beta 1 and ran pgbench against it as follows:

./pgbench -h windows.box -U postgres -P postgres bench -c 10 -t 10000

I encountered the following errors:

Client 0 aborted in state 3: ERROR: could not access status of
transaction 24538
DETAIL: could not open file "C:/Program
Files/PostgreSQL/8.0-beta1-2/data/pg_subtrans/0000": Invalid argument
Client 6 aborted in state 3: ERROR: could not access status of
transaction 12277
DETAIL: could not open file "C:/Program
Files/PostgreSQL/8.0-beta1-2/data/pg_subtrans/0000": Invalid argument

I ran pgbench a few times and continued to get errors.

I then ran vacuum to see if I'd have issues access each tuple. Again, the
same error:

bench=# vacuum verbose accounts;
INFO: vacuuming "public.accounts"
ERROR: could not access status of transaction 98865
DETAIL: could not open file "C:/Program
Files/PostgreSQL/8.0-beta1-2/data/pg_subtrans/0001": Invalid argument

I've confirmed that both files 0000 and 0001 exist. They are being used
(we wouldn't be able to process any queries without it, right?).

Is it possible that I am hitting some file limits in windows? Or is it
possibly a hardware issue? (I don't have another machine to test on)

Thanks,

Gavin

Browse pgsql-hackers-win32 by date

  From Date Subject
Next Message Fabien COELHO 2004-09-01 08:31:11 open item: tablespace handing in pg_dump/pg_restore
Previous Message Tom Lane 2004-09-01 04:58:38 Re: Win32 VACUUM VERBOSE