Open Replicator
by Jesse on Sep.09, 2008, under Open Replicator, Replication
Pretty cool toy. Essentially it’s SanCopy for Symmetrix.
So a customer is particularly reticent about upgrades, to the point that they are going from an 8830 directly to a DMX-4. This is an easy data migration, but they have a handful of test/dev servers that are going to stay on the 8830 for the time being.
So enter Open Replicator. It actually turned out to be easier than I expected. Zone the FA’s, mask the source FA’s to the target luns, create the device pairing file, and issue a couple of simple commands on the “Control” (Usually source) side to push the data across.
The fun part is of course that, like SanCopy, if you mount a filesystem on the target side, you have to issue a full sync. Because after all, SanCopy isn’t like SRDF, there is no persistent cache table tracking changed tracks on the target side. When the track changes on the target device, the only way it’s going to get copied is if it’s changed on the source side as well.
Very cool. I always like playing with new toys.
September 10th, 2008 on 6:33 am
OR has many, many neat capabilities…were you using OR in Live Migrate mode? You should be able to connect the new LUNs on the new array to the old (source) volumes, move the hosts to the new (target) LUNs, initiate a hot pull, and carry on from there without further concern. Any blocks requested by the hosts that aren’t yet in the target will be pulled out-of-order to satisfy the requests. And there is also a sync option that will mirror any writes to the target back to the old source, if you so desire, so that both are kept up to date during the migration.
September 10th, 2008 on 7:37 pm
Actually we’re doing it backwards. Production was migrated from the 8830 to the DMX-4 using RDF, that went smoothly.
However the test and dev systems were left on the 8830. (Customer apparently wasn’t ready to buy enough storage for their whole environment.)
Now the downside of course is that while you can RDF from an 8830 to a DMX, it’s a one-way migration. You can’t RDF back.
So they need a way to move copies back from the DMX for development and testing. It’s either the old fashioned rsync (doesn’t work too well when the database is writing to raw disks) or Open Replicator.
What I’m thinking is we’ll replicate from the production disks to a “Gold” copy on the 8830. Then using TF/Clone we’ll make whatever Test/Dev copies we need to from there. That allows us to preserve the “Gold” copy (as well as the fact that by preserving the gold copy, you preserve the ability to do differential incremental updates) and do multiple updates from a single point in time copy.
September 11th, 2008 on 6:34 am
Ah – now I understand. Makes sense…