On 12/1/2008 6:21 PM, Troy Wolf wrote:
> First, a question that I know has been asked before, but after
> grepping and perusing months of the maillist archives, I did not find
> the answer: Why is it required to use a Subscriber to generate log
> files? Why can't the Origin slon process be started with the option to
> generate logs? It just so happens that in my case, I have an Origin
> and a normal Subscriber and a third node that I replicate via Log
> Shipping. However, I imagine lots of people would have a need for a
> single Origin and a single, WAN-connected node they want to replicate
> via Log Shipping.
> 
> And now onto the issue at hand...
> 
> PostgreSQL 8.2 on SUSE ES 10
> Slony1-1.2.12
> 
> The documentation at http://slony.info/documentation/logshipping.html states:
> ----------------------------------------------------------------------------------------------------------------------------
> 
> 14.2. What takes place when a FAILOVER/ MOVE SET takes place?
> 
> Nothing special. So long as the archiving node remains a subscriber,
> it will continue to generate logs.
> 
> If the archiving node becomes the origin, on the other hand, it will
> continue to generate logs.
> 
> ----------------------------------------------------------------------------------------------------------------------------
> Hmmm..."If the archiving node becomes the origin, on the other hand,
> it will continue to generate logs.". Yes, indeed, it continues to
> generate logs. Unfortunately, the logs do not include any of the
> changes applied to the new Origin!

Well, it certainly is a bug that it continues to generate logs ... we 
will see to fix it and stop it from creating useless files.

The reason is the same as for not being able to create such a setup in 
the first place. Log shipping is implemented as a byproduct of the 
normal replication process done by a subscriber. That includes keeping 
track of what has been replicated thus far. If you want an origin to 
generate the log, it would need another thread, eventually kicked off by 
the sync thread, that queries its own database the same way an actual 
subscriber does and keeps track of what has been written locally.

It could of course be implemented this way, but so far nobody had 
"enough" need for it to provide a patch that implements it.


Jan


> 
> We are currently building a test system to reproduce this behaviour,
> but twice now, we've seen that when we performed a Switchover--that is
> swapped roles between our Origin and Subscriber so the node that was
> the Subscriber and also the node that generates the log files is now
> the Origin--the updates to the new Origin do not make it into the log
> files. The log files keep generating just dandy, but they only contain
> the normal Slony-generated shell (the transaction start and commit,
> etc.) and none of the actual SQL modifications.
> 
> Note that the node that had previously been the Origin that is now the
> normal Subscriber does get the updates. That is, the normal Slony
> replication is working. Only Log Shipping has broken after the
> Switchover.
> 
> BUG? Solved in a post 1.2.12 version?
> 
> Thank you,
> Troy
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general


-- 
Anyone who trades liberty for security deserves neither
liberty nor security. -- Benjamin Franklin