[
https://issues.apache.org/jira/browse/KNOX-481?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14329230#comment-14329230
]
Sumit Gupta commented on KNOX-481:
----------------------------------
I took care of most of the comments above. Here are answers to some things that
didn't go in a commit.
*gateway-test/src/test/java/org/apache/hadoop/gateway/deploy/DeploymnetFactoryFuncTest.java
testSimpleTopology()
**Are there any implications to the change of the role name for webhdfs
dispatch from hdfs to webhdfs?
**Just want to make sure there wasn't any special code in HdfsDisptach since
the tests indicates GatewayDispatchFilter is now used.
{quote}
This seems to not effect anything. Especially now that the contributor is gone,
this just becomes the filter's name in gateway.xml, which now matches the
service name.
{quote}
*gateway-provider-ha/src/main/java/org/pache/hadoop/gateway/ha/provider/HaServletContextListener.java
What motivated the change of the PROVIDER_ATTRIBUTE_NAME? Just curious.
*
gateway-service-webhdfs/src/main/java/org/apache/hadoop/gateway/hdfs/dispatch/WebHdfsHaHttpClientDispatch.java
Did you have to do something special to get @Configure setHaProvider to work? I
thought the config injector stuff only worked with things that could be coerced
to/from strings.
{quote}
Both these questions are related. I changed the config injector to work with
objects and that's how the HaProvider instance can get injected.
{quote}
*gateway-spi/src/amin/java/org/apache/hadoop/gateway/topology/Topology.java
**Is there ever a case where you ned to find a service by just role and name
and not version?
*gateway-spi/src/test/java/org/apache/hadoop/gateway/topology/xml/TopologyRulesModuleTests.java
**Same question as above since you don't test for this.
I added tests for this. This was a good catch although, other than test code,
no one uses this method yet. The caveat in the current implementation is that
the service *must* be looked up exactly as it has been defined in the topology
i.e. if it has name, role and version, you must provide all. I don't know if
this is a bad thing yet.
*gateway-server/src/main/java/org/apache/hadoop/gateway/deploy/impl/ServiceDefinitionDeploymentContributor.java
addDispatchFilterForClass
**The special case for Hive being applied to all dispatches isn't great.
{quote} I have added a todo in the code for this. I agree this is not great. I
was thinking of adding params to the service def or some other way to provide
config to dispatch. We should discuss this and I can create another JIRA for
enhancing this functionality.
{quote}
*
gateway-server/src/main/java/org/apche/hadoop/gateway/topology/builder/property/interpreter/ServicePropertyInterpreter.java
**Does the TODO:sumit need to be resolved before commit?
{quote} I left this in on purpose. I don't think we use the Ambari format. If
we need to, then we need to update this class appropriately.
{quote}
> Support configuration driven REST API integration (aka Stacks)
> --------------------------------------------------------------
>
> Key: KNOX-481
> URL: https://issues.apache.org/jira/browse/KNOX-481
> Project: Apache Knox
> Issue Type: New Feature
> Components: Server
> Affects Versions: 0.6.0
> Reporter: Sumit Gupta
> Assignee: Sumit Gupta
> Fix For: 0.6.0
>
>
> The gateway server has a framework that allows for plugging in new
> 'services'. The framework requires code to be written for every new service
> that is added in. We need a way to make adding new services easy and make it
> possible to add a new service without writing any code (using configuration
> instead) while leaving open the flexibility of being able to write code if
> necessary.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
