[
https://issues.apache.org/jira/browse/NIFI-7799?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17198048#comment-17198048
]
ASF subversion and git services commented on NIFI-7799:
-------------------------------------------------------
Commit e952a76e4f9eba5e668ff06af5ff95554b851acf in nifi's branch
refs/heads/support/nifi-1.12.x from Matt Burgess
[ https://gitbox.apache.org/repos/asf?p=nifi.git;h=e952a76 ]
NIFI-7799: Relogin with Kerberos on connect exception in DBCPConnectionPool
(#4519)
> Have DBCPConnectionPool relogin to Kerberos on connection failure
> -----------------------------------------------------------------
>
> Key: NIFI-7799
> URL: https://issues.apache.org/jira/browse/NIFI-7799
> Project: Apache NiFi
> Issue Type: Improvement
> Components: Extensions
> Reporter: Matt Burgess
> Assignee: Matt Burgess
> Priority: Major
> Fix For: 1.13.0, 1.12.1
>
> Time Spent: 2h 10m
> Remaining Estimate: 0h
>
> An issue can arise (for example when using a Load Balancer and Kerberos)
> where a DBCPConnectionPool can return a connection for which the kerberos
> ticket is not valid for the target node. One example is the case of a load
> balancer in front of an Impala configured for High Availability. When doing a
> rolling restart of Impala, the connection returned points at the load
> balancer but refers to a Kerberos ticket for the Impala node that was
> stopped, causing the controller service to never get successfully connected.
> The DBCPConnectionPool must be restarted in that case.
> Instead, we could catch a connect exception and relogin with Kerberos at that
> point. There may still be connection failures until the node fails over but
> at that point the Kerberos ticket should be good and Impala should return a
> valid connection via the load balancer.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
