[wildfly-dev] Migration management operation - open questions

[Brian Stansberry]

On 8/13/15 10:03 AM, Brian Stansberry wrote:
> On 8/13/15 3:11 AM, Stuart Douglas wrote:
>>
>>
>> On Thu, 13 Aug 2015 at 16:27 Ladislav Thon <lthon at redhat.com
>> <mailto:lthon at redhat.com>> wrote:
>>
>>
>>       > The approach I took was to add the warnings to the describe-migration
>>       > op, and when running the migrate op make the describe-migration
>>       > operation the first step in the composite, so the output looks like
>>       > this: http://pastebin.com/B01KNAHX
>>
>>      The fact that it's a composite operation makes the output a bit cryptic
>>      (step-X?), but otherwise I think it's actually good.
>>
>>      I guess that only handling the warnings in :describe-migration makes the
>>      implementation easier (not sure?) and printing :describe-migration as a
>>      part of :migrate makes sense too, precisely because it shows the steps
>>      that are performed.
>>
>>      So my only objection would be those cryptic "step-X" names, but I guess
>>      that we can live with that (if all :migrate operations took the same
>>      approach; inconsistency wouldn't really help here).
>>
>>
>> That is because this is implemented as a composite operation, AFAIK
>> there is lots of special handing for composite ops, so I don't think we
>> want to be implementing our own one with different output just for this
>> case.
>>
>> I could be wrong though, this is not really my area, Brian will know more.
>>
>
> In the success case, I don't see the point of outputting "step-2" =>
> {"outcome" => "success"}, "step-3" => {"outcome" => "success"}, etc.
>
> In the failure case though, having an intelligible failure-description
> is important. And the composite op handler may be getting in your way there
>
> Instead of writing more text, I'll write some code to illustrate some
> thoughts and post back later.
>

Here's what I was thinking.

[1] is a change to break out the step registration stuff 
CompositeOperationHandler does into a utility method, with a couple 
other variants added that are better suited for use by other OSHs.

Basically, it lets you pass in a Map of ops and a holder map of response 
nodes, and it will add the steps and populate the response node map. 
(You can optionally pre-populate the response node map. There's also a 
variant that uses lists instead of maps in case that's more convenient.)

Using that, the response nodes for the added ops don't have to directly 
appear in the overall operation response. Instead the migrate op handler 
can hold a ref to them and in its ResultHandler use them to format an 
understandable response.

The top two commits in [2] are an example tweak to the current 
WebMigrationOp in master to illustrate use of this.

[1] https://github.com/wildfly/wildfly-core/pull/956

[2] https://github.com/bstansberry/wildfly/tree/help-migrate-multistep

>> Stuart
>>
>>
>>      LT
>>      _______________________________________________
>>      wildfly-dev mailing list
>>      wildfly-dev at lists.jboss.org <mailto:wildfly-dev at lists.jboss.org>
>>      https://lists.jboss.org/mailman/listinfo/wildfly-dev
>>
>>
>>
>> _______________________________________________
>> wildfly-dev mailing list
>> wildfly-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/wildfly-dev
>>
>
>


-- 
Brian Stansberry
Senior Principal Software Engineer
JBoss by Red Hat