Robert Levas created AMBARI-11396:
-------------------------------------

             Summary: Kerberos: UI shows Kerberize Cluster step as failed with 
a retry button, but the backend keeps moving forward to Kerberize the cluster
                 Key: AMBARI-11396
                 URL: https://issues.apache.org/jira/browse/AMBARI-11396
             Project: Ambari
          Issue Type: Bug
          Components: ambari-server
    Affects Versions: 2.0.0
            Reporter: Robert Levas
            Assignee: Robert Levas
             Fix For: 2.1.0


Precondition: On the cluster where this was observed, there were previous 
attempts to kerberize the cluster.

STR:
Go through the Enable Kerberos Wizard.
The only non-default option taken was to not manage krb5.conf (presented on the 
second page of the wizard).

Chrome developer tool shows that there was a POST on 
/api/v1/clusters/woah/artifacts/kerberos_descriptor failing with 409.  

{code}
{
"status" : 409,
"message" : "Attempted to create an artifact which already exists,
artifact_name='kerberos_descriptor',
foreign_keys='{Artifacts/cluster_name=woah}'"
}
{code}

It doesn't seem like this is the cause of the issue (though we need to 
investigate).

The UI keeps showing a spinner for several minutes, then shows a failure.
This is because a call to PUT on the cluster resource to set security_type 
takes more than 3 minutes, and the browser aborts the request.
However, the backend kept moving forward to Kerberize the cluster 
(ambari-server.log was being tailed to check on progress).
After verifying that all principals and keytabs were generated/distributed, the 
wizard was closed (the last step of the wizard is to start all services and run 
service checks, but this was skipped because the previous step failed.)
The cluster was in fact successfully Kerberized.

*Note:* The condition is likely to have occurred due to a timeout related to 
the number of hosts and services in the cluster.  The preparation phase of 
enabling Kerberos is performed within the handler of the relavant API call. 
Most of this work should be moved out to a stage which is handled 
asynchronously.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to