Clients impacted by librdkafka bug in 2.2.0

[rwkarg]

Description

After maintenance operations on clusters (ex. moving nodes) or other communication interruptions, some clients will spam OffsetForLeaderEpoch and Metadata calls until the process is restarted.

This appears to be due to the below issue in librdkafka that is fixed but not yet released:
confluentinc/librdkafka#4425

How to reproduce

1. Have client application connect to a broker
2. Perform maintenance operation (add/remove nodes) or interrupt network communication
3. Occasionally the client will start calling OffsetForLeaderEpoch and Metadata thousands of times a second

[rwkarg]

Is there somewhere that a roadmap can be followed for when new versions of librdkafka and confluent-kafka-dotnet are planned to be available and when they're released?

[anchitj]

Hi, we are planning to release version 2.3.0 this week, which should address this fix. We already have a release candidate (RC) available with this fix if you'd like to test it. You can find it here: https://www.nuget.org/packages/Confluent.Kafka/2.3.0-RC3.

[rwkarg]

Where are those RC releases announced? I don't see it on the Releases page: https://github.com/confluentinc/confluent-kafka-dotnet/releases

[anchitj]

Since it's a RC, it's not announced there. We will announce once we do the final release for 2.3.0

[anchitj]

2.3.0 is released now which has this fixed. Closing. Feel free to reopen if the issue still persists.