[chef] Re: [chef-dev] Re: Re: Re: The future of the database and application cookbooks

["Jason J. W. Williams" < >]

Hi Noah,

On Tue, Aug 30, 2011 at 11:24 PM, Noah Kantrowitz 
<
 >
 wrote:
> Can you point out anything specific that is incompatible with what I've 
> said? The general structure of having callbacks for "create server", 
> "attach slave", "create database", "create user", etc seems suitably 
> generic to me as no assumption about the internals of those is made and 
> most of the database DSL is just collection the params into hashes to send 
> to the backend. If the end point of this is that it can only work for 
> MySQL, Postgres, and SQL Server that still seems pretty useful to me at 
> least.

I'm less qualified to speak on the overlap between MySQL and Postgres
for concepts like users and replication, just because I don't have
much experience with Postgres, but my gut tells me there's probably
enough difference in things like "master replication log name" and
"master replication log position" that the DSL will become non-trivial
quickly to cover all the bases. With systems like Riak, as Denis noted
the concept of a slave doesn't really exist except in the enterprise
edition, and the concept of a user is pretty much non-existent.

We've tended to address "clusters" of MySQL pairs with data bags that
determine who is the master or not, rather than roles for masters and
slaves. Mostly, because for us those two types of MySQL servers have
very similar configurations. The differences are in what the cluster
is responsible for and who is the master, and we prefer to have that
all in once place in a single databag.

-J