-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/31231/
-----------------------------------------------------------
Review request for Ambari, John Speidel and Robert Levas.
Bugs: AMBARI-9733
https://issues.apache.org/jira/browse/AMBARI-9733
Repository: ambari
Description
-------
This patch resolves AMBARI-9733.
Some errors in the BlueprintConfigurationProcessor were
causing cluster deployments to fail with Blueprints
that were exported from HDFS HA clusters.
This patch addresses the issue by:
- Adding some custom code to check for the case of
properties associated with the "SECONDARY_NAMESERVER"
being present in the stack during a cluster update. A
Blueprint exported from a running HA cluster will not
include this configuration, since the Secondary NameServer
is not present in an HA cluster. The Blueprint config
processor will now detect the case of a property
related to the "SECONDARY_NAMESERVER", and will just
return the original value of the property during processing.
- Adding some custom code to handle the formatting of
the "hbase.rootdir" property in the "hbase-site"
configuration file. This property can have a special
meaning in an HA scenario, and the URL in the
property value will typically refer to a logical
nameservice, rather than a host name. The
Blueprint configuration processor now detects
this situation in an HA environment, and will
leave the property unchanged, which is the correct
behavior for an HA cluster.
- Adds unit test assertions to existing tests to
verify these fixes.
Diffs
-----
ambari-server/src/main/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessor.java
d2af1d7
ambari-server/src/test/java/org/apache/ambari/server/controller/internal/BlueprintConfigurationProcessorTest.java
cac1602
Diff: https://reviews.apache.org/r/31231/diff/
Testing
-------
1. Ran the ambari-server unit tests (all passing).
2. Manually verified that a 3-node HDFS HA cluster can be started with a
Blueprint that was exported from a running HDFS HA cluster (HDFS and Yarn).
3. Manually verified that a 3-node HDFS HA cluster can be started with a
Blueprint that was exported from a running HDFS HA cluster with additional
components that reference HDFS (HBase, Hive, Storm, etc).
Thanks,
Robert Nettleton
