Jialiang Cai created AMBARI-26656:
-------------------------------------

             Summary: Implement isolated multi-cluster management and 
recoverable HBase dependencies
                 Key: AMBARI-26656
                 URL: https://issues.apache.org/jira/browse/AMBARI-26656
             Project: Ambari
          Issue Type: New Feature
            Reporter: Jialiang Cai


Ambari needs one console to manage independently owned clusters without leaking 
host, request/task, configuration or UI state between them. Cluster selection 
must preserve the current Dashboard, while Admin provides a clear cluster 
overview, creation, host allocation, permissions and repository navigation.

This new submission consolidates the implementation and review follow-up 
developed under AMBARI-26654. It includes stable cluster ownership across 
server/agent/REST/STOMP boundaries; durable managed HBase dependency 
deployments with binding, operation, credential and task lineage; common 
dependency constraints for Blueprint and Add Service; and recoverable React 
installation/navigation flows.

Acceptance covered a native RPM deployment with six registered nodes: the first 
cluster owns three nodes, the other three are initially unassigned, then API 
provisioning creates the second cluster and adds HBase to both. Installed HBase 
SDK operations verify distinct HDFS namespaces and local ZooKeeper ownership. 
Scoped-user requests, duplicate host attachment and foreign request lookup are 
rejected. Real lifecycle testing found and corrected Registry DNS supervisor 
shutdown and command-scoped HBase hook initialization defects.

Earlier runtime evidence also covers HBase using another cluster's managed 
HDFS/ZooKeeper, actual scoped STOMP events and browser route rejection. 
Production source, focused test commands, preserved failure/recovery receipts 
and remaining acceptance gates are recorded in 
docs/design/multi-cluster-review-and-remediation.md.

Real KDC callbacks, the complete managed provider lifecycle/retention matrix, 
all crash/response-loss boundaries and full broker revocation coverage remain 
open. This issue does not claim those unexecuted scenarios passed.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to