Messing around with MACH11

I recently had the pleasure of playing around with MACH11 in a demo environment. In much the same way as onbar is trivial to set up once you have a storage manager going, once you've got all the background hoo-hah of building instances done, it's amazingly simple to build a MACH11 cluster!

You need to tell the primary that it's the primary for each kind of replication you're going to use (with onmode -d commands. In my case it was HDR, SDS and RSS -- going for broke.

• You need to set up some non-shared disk for SDS.


• Take a backup of the primary and do a physical restore on each secondary in turn.


• You then tell each secondary which kind of secondary it is and where the primary is, also using a single onmode -d command on each secondary.

And that, my friends, was that.

We set up an HDR pair, one SDS secondary and three RSS secondaries in the space of about five minutes!

Pulling the plug on the primary and then "promoting" the SDS secondary to primary was all of another onmode -d command. It was kind of creepy watching all the secondaries detect the new primary and pick up the pieces automagically. Yes, I know it was largely a canned demo and there wasn't any real load or anything, but they were all separate instances and it really did seem to work as advertised. Colour me impressed! :o)

Unfortunately, after all that, I got yanked back into the real world due to another alleged database performance crisis before I could test out redirected writes or the connection manager. :o(

Needless to say, it was nothing of the kind, it turned out to be some spanner in the ops team using RAID-5 on the storage manager's disks (yes, really!) causing a knock-on effect when backups were running.

I'll just have to try again.

Thanks to my man at IBM for setting up the demo ...