Adding new nodes to scidb cluster


#1

I have installed scidb cluster using cluster_install script.
everything is working but i need to expand cluster to include more nodes.
i don’t want to run cluster_install again as it will wipe out my data.

is there a way to add node incrementally without wiping out existing data?

Thanks
Stanislav


Growing cluster with new nodes
#2

Aaaah! Elasticity.

  1. We’re working hard to add elasticity to the product, and to make its use as easy as possible. The engineering isn’t easy. And we’re very conscious of the need to get the testing right. It will take several more months. Hang in there …

  2. In the meantime! The process you need to use at the moment (14.12) is pretty tedious. I’m assuming that what you want to do is to add extra instances.

2.1 First, you need to unload (that is, save ( array, ‘per-instance-file’, -1, ‘opaque’) ) all of the arrays you have. That little example uses the parallel opaque mode, because it’s much more efficient than binary or text. (It’s also a good idea to move the data out of the per-instance data directories and into another location, just in case you fat-finger something and blow away the data storage.)
2.2 You’ll need to create the new installation (new config.ini entries), initialize it, and bring it up.
2.3 Then, create links from the per-instance opaque saves you made previously; to the ‘per-instance-file’ file in the data directory. If you’re going from (say) 4 instances, to 8 instances, move the opaque data files into just 4 instances of the new installation. Then you’ll be able to load the data with 4-way parallelism.

As part of the load ( array, ‘per-instance-file’, -1, ‘opaque’), we will re-distribute the data over the 8 new instances.

Then … delete the old instance, and the external files. And congratulations! You’ve moved from 4 to 8 instances.