Exchange 2007 clustering advice

I appreciate it’s been a while since I blogged last – a combination of "not much to talk about, really" with even more "no time to talk about it"… 🙁

Anyway, a few questions came in the other day from a reader:

SCR and CCR seems to work with SAN and DAS. When DAS (direct attached or local storage) is used, and it most probably it’s attached to the Active node, how does the Passive node function if it hasn’t got connection to the DAS/Local storage of the Active node?

In CCR, it’s important to realise that the passive node has its *own* set of disks, which contain its *own* copy of the data – doesn’t really matter if they are DAS or SAN disks (at least not conceptually). So, in a CCR failover scenario, the (as was) passive node switches to being the active node and uses its own copy of the database (which by now becomes the main one). SCR is different in the way failover happens, but in principle it’s similar – the secondary copy of the data is brought online and takes over servicing the clients, but using its own copy of their database.

-Some clients are indicating that having CCR or SCR one wouldn’t have a need for Backup of mailbox servers. Do you have any comments?

Absolutely not. That’s like saying, because my car has an airbag, I don’t need to wear a seatbelt. Check out the High Availability Strategies section of the Exchange documentation for more detail on the options.

Having CCR gives you the ability to fail over in effectively real time, for the purposes of planned maintenance or after an unexpected failure. SCR adds the possibility of having another replica of the data, potentially in a different location, which can be brought online through a manual recovery process (whereas CCR will bring the data back automatically, since it’s part of a cluster).

Backup is still important (What happens if you lose all servers? What about long-term archival of data?) There’s always the possibility that databases could be corrupted or infected in some way, and if that happened, the replica(s) of the databases would also likely suffer the same fate … so taking regular backups would give you the ability to roll back to earlier versions of the database.

There’s always the scenario where users delete some information that needs to be brought back sometime in the future – there are various options around item recovery with Exchange 2007, but if it was deleted (say) a year ago, then you’d be looking at a backup as the means of recovery.

Data Protection Manager would be worth looking into, to help with backup requirements – it allows you to take regular snapshots of a running server, which can later be spooled out to offline storage.

– In SCR, is there a bandwidth utilization estimate used for replicating the Active and Standby/passive node? I understand that in CCR and SCR the log sizes are reduced to 1MB from standard 5MB though.

The log files in Exchange 2007 are reduced from 5Mb to 1Mb anyway – partly because of CCR and LCR (and later SCR), but even if you don’t configure any of the replication technology, you’ll still be on 1Mb logs.

As far as how much bandwidth you’re going to need between nodes for the purposes of replication, well that depends – if your servers are very busy, then they’ll obviously need to shift more data, and latency will come into play.

There is a detailed section in the Exchange TechCenter online documentation which covers planning for replication at a hardware, software configuration and network level.

Leave a Reply

Your email address will not be published. Required fields are marked *