Stuck Consumer (franz-go) when Follower Fetching enabled

Last updated: August 10, 2026

Issue Description

A follower fetching issue was reported where attempting to fetch older data compared to what was stored locally and remotely caused franz-go based consumer to become stuck. 

Symptoms

  1. Franz-go consumer becomes stuck without any client-side errors.
  2. Redpanda logs show warnings similar to:
WARN 2024-12-12 13:07:09,123 [shard 6:fetc] kafka - replicated_partition.cc:581 - ntp {kafka/live_trades/44}: fetch offset out of range on follower, requested: 5606111, partition start offset: 5619393, high watermark: 5647998, leader high watermark: 5647998, log end offset: 5647998, ec: { error_code: offset_out_of_range [1] }
 

Workaround

Temporarily disabling follower fetching resolved the consumer stuck issue. For instructions on how to disable follower fetching, refer to the Follower Fetching documentation.

Root Cause

The root cause appears to be related to a discrepancy between the data retention on partition leaders and followers, combined with the leader forwarding read requests to followers that don't have the requested data.
 
  1. Partition leader stores data for longer than the configured local retention period, while followers do not.
  2. Partition leader forwards read requests to a follower, causing warnings and consumer stuck issues.

Resolution

As of the time this KB  was written (3/25) the following actions were taken:
  1. For the franz-go client, a bug was logged: franz-go issue #885 / PR #886 and was resolved in release 1.18.1
  2. An improvement was logged for Redpanda: Redpanda PR #24781 which has been merged and is currently expected in the 25.1.1 release.

If you need additional assistance please reach out to Redpanda Support