[AMBARI-22805] Blueprints do not handle some failures properly

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 3.0.0, 2.6.2
• Component/s: ambari-server
• Labels: blueprints pull-request-available
• Assignee: Attila Doroszlai
• Reporter: Attila Doroszlai
• Created: 17/Jan/18 20:46
• Updated: 25/Jan/18 08:06
• Resolved: 25/Jan/18 07:26

Description

Failures in the cluster configuration task and topology host tasks during blueprint cluster deployment or upscaling are not visible via request status. The logical request stays PENDING even after Ambari Server gave up retrying. Both the fact that it no longer makes progress and any reason of the failure can be seen only in ambari-server.log.

Some ways to reproduce (all via blueprints):

• Create cluster with ZooKeeper and HDFS, but omit the NAMENODE component
• Create a secure cluster with wrong Kerberos credentials
• Create a secure cluster without storing Kerberos credentials, restart Ambari Server, add a new node

Issue Links

• links to: GitHub Pull Request #146
• links to: GitHub Pull Request #159 (trunk)
• links to: GitHub Pull Request #179
• links to: GitHub Pull Request #185