Sönke Liebau created NIFI-16294:
-----------------------------------
Summary: NPE in SiteToSiteStatusReportingTask when root flow is
unavailable
Key: NIFI-16294
URL: https://issues.apache.org/jira/browse/NIFI-16294
Project: Apache NiFi
Issue Type: Bug
Affects Versions: 2.11.0
Environment: Observed when running 2.9.0 on Kubernetes, but
objectively unrelated to the environment.
Reporter: Sönke Liebau
Assignee: Sönke Liebau
We recently encountered infrequent NullPointerExceptions when using the
SiteToSiteStatusReportingTask in NiFi. We looked at the code and investigated
in the cluster and usually those Exceptions correlated to cluster nodes
restarting and rejoining the cluster.
{code:java}
SiteToSiteStatusReportingTask[id=xxx] Error running task
SiteToSiteStatusReportingTask[id=xxx]: java.lang.NullPointerException
{code}
*Other issues*
NIFI-7105 sounds very similar to this, but is not in fact related.
*Root cause*
We think we found the root cause and a possible fix, I'll open a PR to
facilitate discussion on the code as well soon.
The SiteToSiteStatusReportingTask calls
context.getEventAccess().getControllerStatus() in
[onTrigger|https://github.com/apache/nifi/blob/f5e20e3687f3eee9fecb3a6b59a5b542390616ef/nifi-extension-bundles/nifi-site-to-site-reporting-bundle/nifi-site-to-site-reporting-task/src/main/java/org/apache/nifi/reporting/SiteToSiteStatusReportingTask.java#L126],
which returns
null when the root process group isn't available (e.g. during startup before
the flow is fully initialized, or when nodes join).
The line immediately following that line guards against `null` when retrieving
the name, but [further
down|https://github.com/apache/nifi/blob/f5e20e3687f3eee9fecb3a6b59a5b542390616ef/nifi-extension-bundles/nifi-site-to-site-reporting-bundle/nifi-site-to-site-reporting-task/src/main/java/org/apache/nifi/reporting/SiteToSiteStatusReportingTask.java#L146]
the result of getControllerStatus is passed on directly to deserialization,
which throws an Exception in the case of `null`.
*Fix*
For a potential fix I looked at the
[SiteToSiteMetricsReportingTask|https://github.com/apache/nifi/blob/f5e20e3687f3eee9fecb3a6b59a5b542390616ef/nifi-extension-bundles/nifi-site-to-site-reporting-bundle/nifi-site-to-site-reporting-task/src/main/java/org/apache/nifi/reporting/SiteToSiteMetricsReportingTask.java#L160]
, which already guards against `null` and skips execution in that case
Something similar should be done in the StatusReporting task to skip the
trigger, when the status cannot be retrieved.
I'm happy to open a PR to discuss directly on the code.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
