It depends a little on the nature of the change, but you need some coordination between the schema change and your code. e.g. add new column, change code to write to it or add new column, change code to use new column and not old column, remove old column. 

Cheers

-----------------
Aaron Morton
New Zealand
@aaronmorton

Co-Founder & Principal Consultant
Apache Cassandra Consulting
http://www.thelastpickle.com

On 19/12/2013, at 3:02 am, Ben Hood <0x6e6562@gmail.com> wrote:

Hi,

I was wondering if anybody knows any best practices of how to apply a
schema migration across a cluster.

I've been reading this article:

http://www.datastax.com/dev/blog/the-schema-management-renaissance

to see what is happening under the covers. However the article doesn't
seem to talk about concurrent write access during the migration
process.

I'm naively assuming that you'd need to block all writes to the
cluster before the migration is started. This is would be firstly
because of potential consistency issues amongst the cluster nodes. But
this would also be because you'd need two versions of your app to
running at the same time.

Does anybody have any experience with doing this kind of thing?

Cheers,

Ben