Hi Rajiv, that is a very good question.
Within a given cluster, Kafka consumers keep track of their offsets in a topic called “__consumer_offsets”. The obvious idea of replicating this topic will not work, because of the different situations, which may cause the offsets to not point to the same messages between the clusters.
Instead, message creation timestamp should be used to find consumer offset before a failure.
Different KIPs (Kafka Improvement Proposal) have been implemented to provide this feature:
KIP-32 – Add timestamps to Kafka message,
KIP-33 – Add a time based log index,
KIP-79 – ListOffsetRequest/ListOffsetResponse v1 and add timestamp search methods to the new consumer,
KIP-122: Add Reset Consumer Group Offsets tooling.
Thus, consumer offset can be reset to the offset, derived from a timestamp before a failure, either from a client application, using the Kafka consumer API, or outside of the application, using Kafka command line tool.

Hi Rajiv, that is a very good question.
Within a given cluster, Kafka consumers keep track of their offsets in a topic called “__consumer_offsets”. The obvious idea of replicating this topic will not work, because of the different situations, which may cause the offsets to not point to the same messages between the clusters.
Instead, message creation timestamp should be used to find consumer offset before a failure.
Different KIPs (Kafka Improvement Proposal) have been implemented to provide this feature:
KIP-32 – Add timestamps to Kafka message,
KIP-33 – Add a time based log index,
KIP-79 – ListOffsetRequest/ListOffsetResponse v1 and add timestamp search methods to the new consumer,
KIP-122: Add Reset Consumer Group Offsets tooling.
Thus, consumer offset can be reset to the offset, derived from a timestamp before a failure, either from a client application, using the Kafka consumer API, or outside of the application, using Kafka command line tool.