Agreed. I was desperate. It turned out that we had a Java server process
that was using the 7.4.5 version of the JDBC driver with
autocommit=false, and Connection.commit() in that version of the driver
does 
"commit; begin;" which obviously leaves the postgres process "<IDLE> in
Transaction" - apparently a well-known issue but I was not aware of it.
THAT made for a fun weekend.....

Thanks.

- DAP

>-----Original Message-----
>From: slony1-general-bounces at gborg.postgresql.org 
>[mailto:slony1-general-bounces at gborg.postgresql.org] On Behalf 
>Of Andrew Sullivan
>Sent: Monday, May 02, 2005 10:38 AM
>To: slony1-general at gborg.postgresql.org
>Subject: Re: [Slony1-general] transaction still in progress
>
>On Sat, Apr 30, 2005 at 03:10:17PM -0400, David Parker wrote:
>> So my question is: is there any way to signal slon to go ahead with 
>> the copy, or to somehow manually update the slony database 
>to indicate 
>> that the transactions are up-to-date? Obviously one doesn't 
>want to go 
>> blindy doing stuff like this, but "faking-out" the slon 
>would be less 
>> harmful in this case than killing postmaster processes....
>
>How can you be sure of that?  (This question is the one which 
>implicitly underlies the decision to rely on xid.)  If people 
>are sitting open in transactions, you can't know that telling 
>slony to go ahead is safe -- they might yet do something which 
>breaks you.
>
>A
>
>--
>Andrew Sullivan  | ajs at crankycanuck.ca
>This work was visionary and imaginative, and goes to show that 
>visionary and imaginative work need not end up well. 
>		--Dennis Ritchie
>_______________________________________________
>Slony1-general mailing list
>Slony1-general at gborg.postgresql.org
>http://gborg.postgresql.org/mailman/listinfo/slony1-general
>