[
https://issues.apache.org/jira/browse/KAFKA-18477?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17931915#comment-17931915
]
Chia-Ping Tsai commented on KAFKA-18477:
----------------------------------------
{quote}
It's unlikely that any topic would only have V0/V1 records since any writes
with IBP >= 3.0 would result in a V2 record being written.
{quote}
I understand it's an edge case, but a topic with only v0/v1 records is still
valid in 4.x, right?
{quote}
latestEpoch(tp) should always be available since we only have V2 message format
in 4.0
{quote}
Regardless of record version, the leaderEpochCache is updated when appending
records. Therefore, the follower fetcher doesn't know the latestEpoch when
starting the fetch thread. However, this looks like a no-op because it
truncates the offset to 0.
{code:java}
[2025-03-03 09:21:43,792] INFO [ReplicaFetcher replicaId=1, leaderId=0,
fetcherId=0] Truncating partition 33opic-0 with TruncationState(offset=0,
completed=true) due to local high watermark 0
(kafka.server.ReplicaFetcherThread)
[2025-03-03 09:21:43,792] INFO [UnifiedLog partition=33opic-0,
dir=/tmp/kafka-logs] Truncating to 0 has no effect as the largest offset in the
log is -1 (kafka.log.UnifiedLog)
{code}
> remove usage of OffsetForLeaderEpochRequest in AbstractFetcherThread
> --------------------------------------------------------------------
>
> Key: KAFKA-18477
> URL: https://issues.apache.org/jira/browse/KAFKA-18477
> Project: Kafka
> Issue Type: Improvement
> Components: core
> Affects Versions: 4.0.0
> Reporter: Jun Rao
> Assignee: 黃竣陽
> Priority: Major
> Fix For: 4.1.0
>
>
> This is because of the base MV in 4.0 is 3.0.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
