Guillaume BARTHE wrote:

> *Does someone knows this?
> Is it normal?
>
> Environment:*
>
>    Linux Debian and fedora core 3
>    postgresql7.4.6
>    slony-1.0.5
>
> *Description:*
>
>    I have a master and a slave.
>    The slave suscribe then master
>    My table in my set is very simple (a integer primary key, b integer)
>    I make a loop process 100000 INSERT(i++,3) in the master table.
>    After few seconds, replication is good.
>    After, i do a delete where b=3, the master table is OK, but
>    replication isn't doing and the slave postmaster goes to 99% cpu.
>    Now, on the slave, if I try to do a delete where b=3, posgresql
>    doesn't answer my self, normally postgresql say
>    Slony-I: Table test is replicated and cannot be modified on a
>    subscriber node.
>    ???

Is it possible that all of those 100K inserts took place as one transaction?

If so, it would be quite unsurprising that:
 a) It would take quite a while for that particular SYNC to process,
 b) The subscriber's postmaster would be extremely busy processing that 
SYNC,
 c) The entry with "b=3" isn't present yet on the subscriber, so an 
attempt to delete it will simply not take place because the entry isn't 
there to be deleted yet.

That all seems self-consistent...

If you wait for the 100K inserts on the subscriber to complete, you 
should see the results you "expected."