[
https://issues.apache.org/jira/browse/NIFI-7438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17143824#comment-17143824
]
Christian Danner commented on NIFI-7438:
----------------------------------------
Sorry for my late reply.
It seems that your recommendation works for us!
It is a little unfortunate that the default settings will result in such a
behavior in our environments. Maybe the note can be clarified here, since our
developers didn't consider using the "Validation Query" since the Note is
warning of performance degregation.
It would be a nice feature in NiFi UI to see the amount of connections a
controller service is having to see the issue first sight.
For completeness I attach this grafana dashboard here so if anyone comes along
the same issue in their performance KPI can compare. Other than that, this
case can be closed in my point of view. [~mattyb149]
!nifi-grafana-dashboard-controller-service-stress.png!
> Controller Service UI hangs since controller service doesn't handle
> reconnections properly
> ------------------------------------------------------------------------------------------
>
> Key: NIFI-7438
> URL: https://issues.apache.org/jira/browse/NIFI-7438
> Project: Apache NiFi
> Issue Type: Bug
> Components: Configuration, Core Framework
> Affects Versions: 1.11.1, 1.11.2, 1.11.3, 1.11.4
> Environment: RHEL 7.6
> NiFi on Docker
> Reporter: Christian Danner
> Priority: Major
> Attachments: image-2020-05-11-13-38-30-457.png,
> nifi-grafana-dashboard-controller-service-stress.png
>
>
> We encounter Problems in showing and managing Controller Services on NiFi
> Servers that run longer than connected systems we use to connect via
> Controller Services. Viewing such Controller Services takes about 1 minute to
> open the view and then shows multiple instances of it (see screenshot).
> e.g. A "DBCPConnectionPool 1.11.4" shows multiple "Supported Controller
> Services" all the "DBCPService 1.11.4 from org.apache.nifi -
> nifi-standard-services-api-nar" (19 occurances) as well as shows multiple
> Groups of "Processors" {color:#172b4d}(19 occurances){color} with the same
> instances (9 processor instances).
>
> In this specific case we have a connection to a database server that is
> shutdown during the night (DEV environment). NiFi Server continues to run
> while database is not reachable.
> We suspect that when the database comes available again, it will load the
> controller service completely while not unloading old sessions. Disabling and
> re-enabling the controller service doesn't solve the issue, only a complete
> NiFI Server restart fixes this issue.
> All related processors also fail to execute defined "ExecuteSQL" Processors
> until we restarted the Controller Services at least. We get a
> "SocketException" since it tries to make an attempt on the old connection
> which is not active anymore. For us it is important to have a "Infinite
> Connection" while interacting with the database ({color:#262626}Max
> Connection Lifetime{color:#003000} = -1).{color}{color}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
