-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Andrew Sullivan a écrit :
> On Tue, Jan 06, 2009 at 07:59:22PM +0800, 滢铖 wrote:
> 
>> always see the process of slonyI show “fetch 100 from LOG”, But I found
>> that no data transmission from the origin to the receiver,and there is no
>> error message in the logs of database
>>
>> Now sl_log_1 has 6442122 rows in it. sl_log_2 has 17681112 rows in it, and
>> they still continue to grow
> 
> It's working.  It's moving 100 rows at a time from the origin to the
> receiver.  You might be able to improve things by increasing the -g
> parameter, but I don't really think that will help you.  The main
> thing I'd say is that you have to wait.
>  
>> I tried to made the following efforts: 
>>
>> 1.restart the slonyI service
>>
>> 2.kill the process
>>
>> However, the problem still exists
> 
> By doing this, you rolled back the work you did on the replica.  I bet
> you need to vacuum -- possibly both systems now.
> 
>> Before problems occur, I have an operation : insert into table select * from
>> temp_table,there are 7000000 rows in the temp_table. This action is related
>> with the problem?
> 
> Yep, that's the problem.  All 7 000 000 rows need to be sent to the
> replica.  In a Slony environment you almost never want to do large
> operations like that, for exactly the reason you're seeing.
> 
> I think if it were me, I'd drop the replica and rebuild it from
> scratch.
> 
> A
> 
> 
> 

Search the mailing list for tips about that issue.
1/ there is some patch to increase the size of the cursor
2/ AFAIR there is some enable_indexscan=off or something like that in the slony
code, disabling this in the code can be usefull too.

- --
Cédric Villemain
Administrateur de Base de Données
Cel: +33 (0)6 74 15 56 53
http://dalibo.com - http://dalibo.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAkljYcQACgkQo/dppWjpEvyK6QCguHgGFO0Q7ofClesOVPEnKDgU
qXgAni13rYADwe0nEllNSffKtTarmx3W
=sdqn
-----END PGP SIGNATURE-----