DavidEnnis-CleverLlamas
commented
1 year ago This is a great improvement to go alongside #124. As an example - working locally as a developer(single node ML and single node NiFi and maybe mix container connectivity into it), then it's actually not difficult to [mis]configure yourself into an error state with an eternal loop of error file. At least a developer can focus on one error and not suddenly thousands of lines in the logs to see what went wrong. This helps lower the bar in terms of faster troubleshooting.
rjrudin
These 2 processors differ from the other processors in that when an exception occurs, they rollback the session. This causes the FlowFile to be retried. In the event of a connection error or misconfiguration, the FlowFile will repeatedly fail, leading to the logging described in #124 .
Both for consistency with the other processors and also for what seems like behavior that NiFi wants, these two processors should be modified to send failed FlowFiles to a "failure" relationship. This does happen already for some failures in
PutMarkLogic, but we should do it for all failures. This gives the user the ability to determine how to handle the relationship and avoids the undesirable behavior reported in #124.