navina commented on code in PR #9713:
URL: https://github.com/apache/pinot/pull/9713#discussion_r1186547382
##########
pinot-plugins/pinot-stream-ingestion/pinot-kinesis/src/main/java/org/apache/pinot/plugin/stream/kinesis/KinesisStreamMetadataProvider.java:
##########
@@ -189,6 +192,24 @@ private boolean
consumedEndOfShard(StreamPartitionMsgOffset startCheckpoint,
}
@Override
- public void close() {
+ public Map<String, PartitionLagState> getCurrentPartitionLagState(
+ Map<String, ConsumerPartitionState> currentPartitionStateMap) {
+ Map<String, PartitionLagState> perPartitionLag = new HashMap<>();
+ for (Map.Entry<String, ConsumerPartitionState> entry:
currentPartitionStateMap.entrySet()) {
+ ConsumerPartitionState partitionState = entry.getValue();
+ // Compute record-availability
+ String recordAvailabilityLag = "UNKNOWN";
+ RowMetadata lastProcessedMessageMetadata =
partitionState.getLastProcessedRowMetadata();
+ if (lastProcessedMessageMetadata != null &&
partitionState.getLastProcessedTimeMs() > 0) {
+ long availabilityLag = partitionState.getLastProcessedTimeMs()
+ - lastProcessedMessageMetadata.getRecordIngestionTimeMs();
+ recordAvailabilityLag = String.valueOf(availabilityLag);
+ }
+ perPartitionLag.put(entry.getKey(), new
KinesisConsumerPartitionLag(recordAvailabilityLag));
Review Comment:
I agree that the kinesis official metric will be the most accurate one. Let
me re-think if this approach makes sense because we now have a mechanism to
compute ingestionDelay in the realtime engine itself based on this apprx
arrival time. That is ideally what we should be using uniformly across all
connectors.
> Also, we should add a javadoc here declaring what happens to lag in case
of a reshard operation.
Good point. When re-shard happens, I am guessing one or more of the
partitions may no longer have any data being ingested and cause this lag to
continuously increase, leading to false alarms. Is my understanding right?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]