Github user MikeThomsen commented on a diff in the pull request: https://github.com/apache/nifi/pull/2101#discussion_r165812983 --- Diff: nifi-nar-bundles/nifi-influxdb-bundle/nifi-influxdb-processors/src/main/java/org/apache/nifi/processors/influxdb/PutInfluxDB.java --- @@ -78,18 +81,33 @@ .addValidator(StandardValidators.NON_EMPTY_VALIDATOR) .build(); + static final Relationship REL_SUCCESS = new Relationship.Builder().name("success") + .description("Successful FlowFiles that are saved to InfluxDB are routed to this relationship").build(); + + static final Relationship REL_FAILURE = new Relationship.Builder().name("failure") + .description("FlowFiles were not saved to InfluxDB are routed to this relationship").build(); + + static final Relationship REL_RETRY = new Relationship.Builder().name("retry") + .description("FlowFiles were not saved to InfluxDB due to retryable exception are routed to this relationship").build(); + + static final Relationship REL_MAX_SIZE_EXCEEDED = new Relationship.Builder().name("failure-max-size") --- End diff -- Something to consider here... With PutHBaseRecord and (still pending merge) DeleteHBaseRow, I had a similar situation. Users could easily chuck several hundred thousand HBase operations at the processor all at once. So what was suggested to me, and I did with both of them, was to break up the incoming flowfile into chunks and then add a "retry.index" attribute to the flowfile if it failed. That way, users could loop REL_RETRY to the processor and get everything ingested. Though that might not apply in this case because InfluxDB doesn't have an ID field that I know of. If you replay an event with the same timestamp, does it overwrite or does it just add a new one?
---