Thread: Backend dies (6.5.3-1) on SELECT from 6 row table

Backend dies (6.5.3-1) on SELECT from 6 row table

From
"Emils Klotins"
Date:
Hello!

Can you please tell me, 
what are the memory requirements for Postgres?

I have a Postgres 6.5.3-1 server , a db with approx 15 tables, 
no table has more than 10-15 fields, mostly text a couple of int.
I have currently no more than 20 rows in any table. Most have 
2-6 rows.Total db directory size is 4MB.

When I try to "SELECT * FROM most of the tables (except for 
ones which have 2 or less rows) , the backend dies:
> smmc=> select * from authors;
> pqReadData() -- backend closed the channel unexpectedly.
>         This probably means the backend terminated abnormally
>         before or while processing the request.
> We have lost the connection to the backend, so further processing
> is impossible.  Terminatin g.

If i try to select specific field (address_id or lastname, 
firstname from the same table - it works).

Postmaster remains and I can connect again immediately.

If I try this thru my webserver interface, I have to restart it else 
it can't reestablish link to db (I'm using Roxen Challenger 
1.3.121). 
However, the web interface error message at least shows that 
Postgres can't allocate memory.

Which is kind of strange. I installed 6.5.3-1 today (RPMs from 
postgresql.org), I am running Red Hat Linux 5.2 (kern 2.0.36) 
on Pentium II 266/ w/ 128MB RAM. Only other significant 
process running is the web server, which is running on other 
machines all right (together with postgres).

The db is a leftover from 6.4.2 I had this morning. I did a 
pg_dumpall -o > db.out and restored it to the new version.


Emils.



Re: [SQL] Backend dies (6.5.3-1) on SELECT from 6 row table

From
Tom Lane
Date:
"Emils Klotins" <emils@mail.usis.bkc.lv> writes:
> When I try to "SELECT * FROM most of the tables (except for 
> ones which have 2 or less rows) , the backend dies:

> If i try to select specific field (address_id or lastname, 
> firstname from the same table - it works).

That's really odd.  I doubt you have a memory problem; corrupted
data or maybe an outright bug of some kind seem more likely.

What error message, if any, shows up in the postmaster log file?

Also, the crashed backend presumably leaves a core file in the database
directory.  Do you know how to use gdb to get a backtrace showing what
the backend was doing?  Briefly it'sgdb path/to/postgres path/to/corebtquit
        regards, tom lane