gharris1727 commented on code in PR #13429: URL: https://github.com/apache/kafka/pull/13429#discussion_r1159083568
########## connect/mirror/src/main/java/org/apache/kafka/connect/mirror/OffsetSyncStore.java: ########## @@ -25,17 +25,39 @@ import org.apache.kafka.common.utils.Utils; import org.apache.kafka.connect.util.KafkaBasedLog; import org.apache.kafka.connect.util.TopicAdmin; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; +import java.util.Arrays; import java.util.Collections; import java.util.Map; import java.util.Optional; import java.util.OptionalLong; import java.util.concurrent.ConcurrentHashMap; -/** Used internally by MirrorMaker. Stores offset syncs and performs offset translation. */ +/** + * Used internally by MirrorMaker. Stores offset syncs and performs offset translation. + * <p>A limited number of offset syncs can be stored per TopicPartition, in a way which provides better translation + * later in the topic, closer to the live end of the topic. + * This maintains the following invariants for each topic-partition in the in-memory sync storage: + * <ul> + * <li>Invariant A: syncs[0] is the latest offset sync from the syncs topic</li> + * <li>Invariant B: For each i,j, i <= j: syncs[j].upstream <= syncs[i].upstream < syncs[j].upstream + 2^j</li> Review Comment: I split this invariant into two parts: the upper bound and lower bound. I also made both of the bounds stronger to promote a more effective distribution of syncs. This means that the old logic of bit comparisons is gone now, and the invariants actually drive the algorithm directly, instead of the bit comparisons happening to satisfy the invariants but not actually having anything to do with them. -- 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: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
