Thread: LockReplace: xid table corrupted

Hi There,

     I have production systems running postgres for
24x7x365. But suddenly it stops serving the requests
but the postmaster does not die. I cant do select or
even say pgsql <database-name> until i kill postmaster
clear shared memory segments and restart it again. for
all the forth coming requests it says "sorry, too many
clients already".

My system config :-

os:- Red Hat linux 6.2
memory:- 1GB
cpu:- dual CPU
database:- Postgres 6.5.2 ( pl dont suggest to upgrade
to 7.0.2. once this problem is solved. then i can do
testing of my app with postgres 7.0.2)

command:- postmaster -i -N 256 -B 512
>>/usr/local/pgsql/server.log 2>&1 &

can anybody tell me why this happens and what the
solution is for this ????

Regards
Rajesh

__________________________________________________
Do You Yahoo!?
Yahoo! Shopping - Thousands of Stores. Millions of Products.
http://shopping.yahoo.com/

Well, not to sound silly but, how many connections exist. Are you out of
connections?

I know that some clients on some OSes fail to free socket connections if
they die. We had this problem with JDBC on Win NT and I've heard of problems
with PHP as well. Our solution was to rewrite part of the clients so that in
all cases they shut down the db connection even if they then die a fiery
death.


--Rainer



> -----Original Message-----
> From: pgsql-admin-owner@postgresql.org
> [mailto:pgsql-admin-owner@postgresql.org]On Behalf Of Dronamraju Rajesh
> Sent: Tuesday, December 12, 2000 2:56 PM
> To: pgsql-admin@postgresql.org; pgsql-bugs@postgresql.org
> Subject: [ADMIN] LockReplace: xid table corrupted
>
>
> Hi There,
>
>      I have production systems running postgres for
> 24x7x365. But suddenly it stops serving the requests
> but the postmaster does not die. I cant do select or
> even say pgsql <database-name> until i kill postmaster
> clear shared memory segments and restart it again. for
> all the forth coming requests it says "sorry, too many
> clients already".
>
> My system config :-
>
> os:- Red Hat linux 6.2
> memory:- 1GB
> cpu:- dual CPU
> database:- Postgres 6.5.2 ( pl dont suggest to upgrade
> to 7.0.2. once this problem is solved. then i can do
> testing of my app with postgres 7.0.2)
>
> command:- postmaster -i -N 256 -B 512
> >>/usr/local/pgsql/server.log 2>&1 &
>
> can anybody tell me why this happens and what the
> solution is for this ????
>
> Regards
> Rajesh
>
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Shopping - Thousands of Stores. Millions of Products.
> http://shopping.yahoo.com/

Dronamraju Rajesh <drajesh80@yahoo.com> writes:
> database:- Postgres 6.5.2 ( pl dont suggest to upgrade
> to 7.0.2. once this problem is solved. then i can do
> testing of my app with postgres 7.0.2)

If you update to 7.0.3, you'll probably find this problem
goes away.  It sounds to me like you are running into a bogus
debug check in the older code that assumed LockRelease should
never need to release more than a thousand locks.

If you don't want to update, try not touching 1000 tables in
the same transaction.  I'm guessing you do a lot of large-object
slinging ... otherwise it's not that easy to get to 1000 ...

            regards, tom lane