vikrambohra commented on code in PR #3517:
URL: https://github.com/apache/gobblin/pull/3517#discussion_r891629386
##########
gobblin-iceberg/src/main/java/org/apache/gobblin/iceberg/writer/IcebergMetadataWriter.java:
##########
@@ -790,28 +789,33 @@ public void flush(String dbName, String tableName) throws
IOException {
Transaction transaction = tableMetadata.transaction.get();
Map<String, String> props = tableMetadata.newProperties.or(
Maps.newHashMap(tableMetadata.lastProperties.or(getIcebergTable(tid).properties())));
+ String topic = props.get(TOPIC_NAME_KEY);
if (tableMetadata.appendFiles.isPresent()) {
tableMetadata.appendFiles.get().commit();
if (tableMetadata.completenessEnabled) {
- String topicName = props.get(TOPIC_NAME_KEY);
- if(topicName == null) {
- log.error(String.format("Not performing audit check. %s is null.
Please set as table property of %s.%s",
- TOPIC_NAME_KEY, dbName, tableName));
- } else {
- long newCompletenessWatermark =
- computeCompletenessWatermark(topicName,
tableMetadata.datePartitions, tableMetadata.prevCompletenessWatermark);
- if(newCompletenessWatermark >
tableMetadata.prevCompletenessWatermark) {
- log.info(String.format("Updating %s for %s.%s to %s",
COMPLETION_WATERMARK_KEY, dbName, tableName, newCompletenessWatermark));
- props.put(COMPLETION_WATERMARK_KEY,
String.valueOf(newCompletenessWatermark));
- props.put(COMPLETION_WATERMARK_TIMEZONE_KEY, this.timeZone);
- tableMetadata.newCompletenessWatermark =
newCompletenessWatermark;
- }
- }
+ checkAndUpdateCompletenessWatermark(tableMetadata, topic,
tableMetadata.datePartitions, props);
}
}
if (tableMetadata.deleteFiles.isPresent()) {
tableMetadata.deleteFiles.get().commit();
}
+ // Check and update completion watermark when there are no files to be
registered, typically for quiet topics
+ // The logic is to check the next window from previous completion
watermark and update the watermark if there are no audit counts
Review Comment:
This case should happen only after a long downtime. Even then watermark
should catchup quickly as we will be flushing every minute. And if there is a
append file operation in between then we would catchup more aggressively
In regular scenario. We will update watermark on the next flush once current
time is the next hour of current watermark.
--
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]
