MSK metric SumOffsetLag - how it works?

Question

Amazon MSK provides a metric named SumOffsetLag, described in the [documentation](https://docs.aws.amazon.com/msk/latest/developerguide/metrics-details.html) as 
> The aggregated offset lag for all the partitions in a topic.

The question is how value of this metric is calculated with respect to the specific Kafka metrics.
It could be either based on 
* [records-lag-max](https://docs.confluent.io/platform/current/kafka/monitoring.html#records-lag-max) metric and then it doesn't take committed offsets into consideration
* [consumer-lag-offsets](https://docs.confluent.io/platform/current/kafka/monitoring.html#consumer-lag-offsets) and then it's based on committed offsets

Answer

For SumOffsetLag, the aggregated offset lag for all the partitions in a topic. It is an offset-based lag and topic level to sum offset lag for all partitions in a topic.

The value of SumOffsetLag metrics with respect to the specific Kafka metrics is calculated based on consumer-lag-offsets and then it’s based on committed offsets

https://docs.confluent.io/platform/current/kafka/monitoring.html#consumer-lag-offsets

MSK metric SumOffsetLag - how it works?

相關內容