Jeff wrote:
>>-----Original Message-----
>>From: Alec.Cawley@stripped [mailto:Alec.Cawley@stripped]
>>Sent: Monday, September 19, 2005 10:10
>>To: jakes.dad@stripped
>>Cc: mysql@stripped
>>Subject: Re: Circular Replication
>>
>>
>>Sid Lane <jakes.dad@stripped> wrote on 19/09/2005 15:02:58:
>>
>>
>>>stupid ?:
>>>
>>>what keeps them from getting caught in a write loop? turning off
>>>log_slave_updates?
>>>
>>>I had never thought of this but is has intriging possibilities...
>>
>>Each update is marked with the unique server id of the server which
>>originated it. When the update returns to its originating
>>server, it is
>>dropped instead of being executed. That is why every server
>>must have a
>>unique id.
>>
>> Alec
>>
>
>
> Actually, isn't it required that you start each server with
> --log_slave_updates?
>
> Or is that only necessary in a replication situation like this with more
> than 2 servers?
>
> A -> B -> C -> A
>
> If it's just:
>
> A -> B -> A
>
> Is it necessary to start the servers with --log_slave_updates?
>
> Thanks,
>
> Jeff
>
>
>
log-slave-updates causes the server to write to it's own binlog any
statements it read from it's master's binlog. This is necessary in
chains of servers (ie, A->B->C->A); in such chains, it is necessary that
all servers have BOTH log-bin and log-slave-updates.
With just 2 servers (A->B->A), you need log-bin, but do not need
log-slave-updates.
Regards,
Devananda vdv
