-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/35087/
-----------------------------------------------------------
Review request for Ambari, John Speidel, Mahadev Konar, Robert Levas, and Sumit
Mohanty.
Bugs: AMBARI-11701
https://issues.apache.org/jira/browse/AMBARI-11701
Repository: ambari
Description
-------
This patch addresses AMBARI-11701.
The Blueprint processor was previously adding the three properties described in
AMBARI-11701 to every HDFS deployment, regardless of whether the cluster was
configured to use NameNode HA. Since these address properties were referring
to single-nodes, the Blueprint processor previously required that users set
these properties in an HA cluster, since it was unclear if these properties
were required.
After some research, it was determined that these three properties are not used
by HDFS in an HA environment. The HA-specific properties for these addresses
are used, rather than the single-node versions.
The Blueprint processor should not require these properties to be set, since
this is confusing and an inconvenience to users.
This patch addresses the problem by:
1. Registering a filter that will remove the three HDFS properties from the
configuration prior to Blueprint processing. This keeps these properties from
being set on the cluster during cluster startup time.
2. Adding unit tests to verify this change.
Diffs
-----
ambari-server/src/main/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessor.java
c6ff56c
ambari-server/src/test/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessorTest.java
d957836
Diff: https://reviews.apache.org/r/35087/diff/
Testing
-------
1. Ran the Blueprint-specific ambari-server unit tests (all passing). I plan to
run the full ambari-server unit test suite prior to merging.
2. Deployed a 3-node HDFS NameNode HA cluster without including these three
properties, and verified that this cluster started properly.
3. Deployed a single-node HDFS non-HA cluster with Hive and HBase, to make sure
the new filter did not adversely affect cluster startup in a non-HA setup.
Verified that the cluster started properly.
Thanks,
Robert Nettleton
