Lotus Notes DR Replication

Hi Experts,
I am trying to establish the most rational method of replicating all user databases from our Live Domino server (v6.5) to our DR server (v8.5.1) at a remote site in a disaster recovery configuration.

Is there a bandwidth efficient method of doing this, much like database journalling as opposed to replication?
Also, is there anything I need to be aware of when setting up the server ID's or in the user configurations (such as enabling roaming, which does not seem to be possible with existing users)?

Who is Participating?
Sjef BosmanConnect With a Mentor Groupware ConsultantCommented:
Isn't it much like a cluster that you intend to set up? What exactly is your DR strategy, do you want users to continue on the other server in case of? If that's the idea, go for clustering indeed.

If you only want a backup server to back up databases, you don't need the cluster. By the way, replication is very efficient, only differences are exchanged between replicas. What's more important: there is always a delay when it comes to replication, the usual is one hour (but it can be more frequent, at the cost of extra network usage of course).
doninjaConnect With a Mentor Commented:
Also to be more efficent make sure that the TCPIP port is set to compress traffic. This will increase CPU overhead slightly but reduce bandwidth.

As Sjef says using replication is efficient and if they can upgrade the live server to R8+ then this will increase as the compression method is better and the mehtod for determining what info is replicated was improved I think around R7 onwards to only replcate field changes rather than complete documents.

For DR strategy for offsite replication wI have previously not useed a cluster as clients would automatically try to use the server if the first server is unavailable for any reason which at first seemed ok but since they did not automatically switch back this was problematical. So the DR server was setup to replicate every 10 minutes during business hours and 2 hours over night when backups where running.

This was disaster recovery not High Availablility solution which would use Clustering.
DR is back to normal operation within a short time (hours) High availablility is continuity of server during normal work hours so normal operation within seconds.

You also need to look at how database changes happen on your sytem to work out best replication times and what is you DR requirement on data continuity. You can also specificy multiple replciation schedules for different databases or folders. Critical databases frequntly and lower priority at night etc.
Replication is indeed very efficient.  Even more so if all the servers involved are of the same version, since there have been significant improvements: streaming replication and DAOS.
DAOS can be a big saver: if an attachment is already present at the replication target server, and it would need to be transferred as part of another document, only the link is replicated, not the attachment.

Another point is that clustering replication tries hard to be up to date immediately. so if you would like users to be able to fail over to the disaster recovery immedeately with as little data loss as possiblem you need to implement clustering.

To be most efficient, you should migrate your production server to 8.5.2 (out next Tuesday, August 24th 2010) and your Dr server too. Then implement DAOS (note: implementation needs backup strategy looked at, see Domino wiki at http://www-10.lotus.com/ldd/dominowiki.nsf/xpViewCategories.xsp?lookupName=Lotus%20Domino contains a wealth of information, including upgrade guides)
Also, you can estimate your storage savings with the DAOS estimator tool, runs with Domino 6 or later: http://www-01.ibm.com/support/docview.wss?rs=463&uid=swg24021920

Stuzza71Author Commented:
Thanks Guys,
This cleared it up for me.
All I need to do now is configure the new server correctly for clustering, essentially we need DR, but also have a tight SLA to keep for the business, so clustering is the way to go.

Worst case, once the primary is back on line, we could drop the Domino service on the HA server to force the clients local again.
All Courses

From novice to tech pro — start learning today.