failover server in case of a critical failure on the primary DB server. I
plan on using transactional replication and immeadiate updating for minimal
latency, but I have a couple of questions about the right setup:
1. Push or pull? Push would allow for 2 phase commit, which seems
desirable. On the other hand, when system 1 fails, don't you need to "break
the link" between the subscriber and publisher (i.e. run sp_dropsubscription)
in order to ensure proper operation?
2. We have some identity columns. I am planning on marking them as NOT FOR
REPLICATION and then in a failover scenario, run DBCC CHECKIDENT('my
table',RESEED,<looked up max from table>) to reseed the identity on the
secondary. Is this an advisable strategy?
I have been looking at other solutions in this area which involve modifying
some of the generated procs and I would like to avoid that so that we can use
as much out of the box as possible (trying not to pretend I can engineer
better than MS
).Thanks,
Tom
Push or pull doesn't really relate to 2PC - the setting really controls who
initiates the distributor->subscriber traffic. If the subscriber fails, the
2PC won't matter because it is in the direction subscriber->distributor. The
other way round and you'll have to remove the subscription.
For the identities, the easiest solution is to set up immediateupdating with
queued updating failover and have a large seed set up.
Cheers,
Paul Ibison SQL Server MVP, www.replicationanswers.com
(recommended sql server 2000 replication book:
http://www.nwsu.com/0974973602p.html)
|||Could you clarify a bit? does subscriber->distributor imply pull whereas
distributor ->subscriber implies push? I am trying to understand when I
would have to remove the subscription.
Thanks,
Tom
"Paul Ibison" wrote:
> Push or pull doesn't really relate to 2PC - the setting really controls who
> initiates the distributor->subscriber traffic. If the subscriber fails, the
> 2PC won't matter because it is in the direction subscriber->distributor. The
> other way round and you'll have to remove the subscription.
> For the identities, the easiest solution is to set up immediateupdating with
> queued updating failover and have a large seed set up.
> Cheers,
> Paul Ibison SQL Server MVP, www.replicationanswers.com
> (recommended sql server 2000 replication book:
> http://www.nwsu.com/0974973602p.html)
>
>
|||The directions I indicated refer to data flow, eg subscriber->distributor in
your case is 2PC, whereas distributor ->subscriber is the PUSH/PULL running
of the distribution agent. I think you're mixing PUSH/PULL with the flow of
data, while the 2 are not really related in that way - it's more which
computer is going to do the synchronization work.
HTH,
Paul Ibison SQL Server MVP, www.replicationanswers.com
(recommended sql server 2000 replication book:
http://www.nwsu.com/0974973602p.html)
