Re: [wildfly-dev] Management operation for legacy subsystem migration
Sent from my iPhone
On Apr 30, 2015, at 9:08 AM, Jeff Mesnil <jmesnil(a)redhat.com>
wrote:
> On 30 Apr 2015, at 04:53, Jason T. Greene <jason.greene(a)redhat.com> wrote:
>
> Instead of hard errors, and manipulating the old model, wouldn't it be better to
just convert and then report back in the response what was dropped?
I’m not sure to understand your remark. Could you clarify?
Sorry I meant wouldn't it be better to skip/ignore configuration that does not have an
equivalent mapping and report just warnings about what doesn't map. My point was that
it's probably easier for the user to fixup the new subsystem then it is to repeatedly
alter the old subsystem until you get something that will convert.
I don’t manipulate the old model. I describe it and use the description to create the
:add operations for the new subsystem.
At first, I was reading the old model, manipulating it to create the new subsystem model
and create a new Resource with that manipulated model.
It turns out to be a bad idea as I was bypassing any validation steps performed by the
:add operations and ended up with an invalid new subsystem (that could not be loaded).
Using the ":describe + new :add operations" approach ensures that my new
subsystem will be valid if my :migrate operation succeeds.
--
Jeff Mesnil
JBoss, a division of Red Hat
http://jmesnil.net/