ConsumerGroup.joinGroup doesn't handle MemberIdRequired (KIP-394)

[erikcw]

Summary

The ConsumerGroup implementation doesn't handle MemberIdRequired (error code 79) per KIP-394. When a broker returns this error with an assigned member ID in the response, the client should retry the JoinGroup request using that member ID. Instead, the current code treats it as a fatal error, discards the member ID, and retries with an empty string, creating an infinite failure loop.

Root cause

In consumergroup.go, the joinGroup method (line ~941) treats any non-zero error code as fatal:

if err == nil && response.ErrorCode != 0 {
    err = Error(response.ErrorCode)
}
if err != nil {
    return "", 0, nil, err  // member ID from response is lost
}

The caller nextGeneration then hits the default case in the error switch, which clears the member ID and retries after a backoff:

default:
    _ = cg.leaveGroup(memberID)
    memberID = ""
    backoff = time.After(cg.config.JoinGroupBackoff)

The member ID returned in the MemberIdRequired response is never preserved.

Expected behavior

Per KIP-394, when the broker returns MemberIdRequired:

1. Extract the MemberID from the JoinGroup response
2. Retry the JoinGroup request with that member ID
3. The second attempt should succeed

How to reproduce

Connect a ConsumerGroup to a Kafka-compatible broker that enforces KIP-394 (e.g., Tansu). The consumer group will fail to join and retry indefinitely with [79] Member ID Required.

Suggested fix

In joinGroup, handle MemberIdRequired by extracting the member ID from the response and retrying:

response, err := conn.joinGroup(request)
if err == nil && response.ErrorCode != 0 {
    if Error(response.ErrorCode) == MemberIDRequired && response.MemberID != "" {
        request.MemberID = response.MemberID
        response, err = conn.joinGroup(request)
        if err == nil && response.ErrorCode != 0 {
            err = Error(response.ErrorCode)
        }
    } else {
        err = Error(response.ErrorCode)
    }
}

Alternatively, handle it in the nextGeneration error switch by not clearing the member ID for this specific error.

References

• KIP-394: Require member.id for initial join group request
• The low-level tests in this repo (joingroup_test.go, syncgroup_test.go) already handle MemberIDRequired correctly with a retry loop, but the high-level ConsumerGroup does not.

[sahasan95]

Hi, I'd like to investigate and fix this. I've read through KIP-394 and understand the expected behaviour. When the broker returns MemberIdRequired, the client should retry the JoinGroup request with the assigned member ID rather than treating it as a fatal error. I'll trace through consumergroup.go to find where the error needs to be handled and submit a PR. Will update here with findings before I start coding.