Github user vanzin commented on a diff in the pull request:

    https://github.com/apache/spark/pull/9232#discussion_r43251011
  
    --- Diff: 
yarn/src/test/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtilSuite.scala 
---
    @@ -245,4 +247,55 @@ class YarnSparkHadoopUtilSuite extends SparkFunSuite 
with Matchers with Logging
           System.clearProperty("SPARK_YARN_MODE")
         }
       }
    +
    +  test("Obtain tokens For HiveMetastore") {
    +    val hadoopConf = new Configuration()
    +    hadoopConf.set("hive.metastore.kerberos.principal", "bob")
    +    // thrift picks up on port 0 and bails out, without trying to talk to 
endpoint
    +    hadoopConf.set("hive.metastore.uris", "http://localhost:0";)
    +    val util = new YarnSparkHadoopUtil
    +    val e = intercept[InvocationTargetException] {
    +      util.obtainTokenForHiveMetastoreInner(hadoopConf, "alice")
    +    }
    +    assertNestedHiveException(e)
    +    // expect exception trapping code to unwind this hive-side exception
    +    assertNestedHiveException(intercept[InvocationTargetException] {
    +      util.obtainTokenForHiveMetastore(hadoopConf)
    +    })
    +  }
    +
    +  def assertNestedHiveException(e: InvocationTargetException): Throwable = 
{
    +    val inner = e.getCause
    +    if (inner == null) {
    +      fail("No inner cause", e)
    +    }
    +    if (!inner.isInstanceOf[HiveException]) {
    +      fail(s"Not a hive exception", inner)
    +    }
    +    inner
    +  }
    +
    +  test("handleTokenIntrospectionFailure") {
    +    val util = new YarnSparkHadoopUtil
    +    // downgraded exceptions
    +    util.handleTokenIntrospectionFailure("hive", new 
ClassNotFoundException("cnfe"))
    --- End diff --
    
    I think that because there's really only one exception that's currently 
interesting, you need more code to implement this "shared policy" approach than 
just catching the one interesting exception in each call site. It's true that 
if you need to modify the policy you'd need you'd need to duplicate code (or 
switch to your current approach), but then do you envision needing to do that? 
What if the policy for each service needs to be different?
    
    Personally I think that the current approach is a little confusing for 
someone reading the code (and inconsistent; for example the current code 
catches `Exception` and then feeds it to a method that matches on `Throwable`), 
and because the policy is so simple, the sharing argument doesn't justify 
making the code harder to follow.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to