Re: [DISCUSS] KIP-630: Kafka Raft Snapshot

[José Armando García Sancio <js...@confluent.io.INVALID>]

During the development of KIP-630 we made some minor changes to the
KIP to better match the implementation details. Here is a summary of
the changes we made to the KIP:

1. Added control records at the begin and end of the snapshots. The
control records are versioned. The snapshot header record includes the
append time of the last record from the log included in the snapshot.
This is useful when determining when to delete a snapshot.
2. The configuration property metadata.snapshot.min.new_records.size
was renamed to metadata.log.max.record.bytes.between.snapshots.
3. The FetchSnapshotRequest schema was changed to include the cluster
id and the current leader epoch. This is used by the leader of the
metadata log to validate that the request matches the current cluster
id and the current leader epoch.
4. The FetchSnapshotResponse schema was changed to include the snapshot id.

KIP-630: https://cwiki.apache.org/confluence/x/exV4CQ

Thanks!
-Jose