Issues with using group instance id with librdkafka

[BalazsHoranyi]

Version & Environment

Redpanda version: (use rpk version):

Latest Helm release (21.1.6)

Please also give versions of other components:

• Kubernetes (use kubectl version):

Server Version: version.Info{Major:"1", Minor:"20", GitVersion:"v1.20.7", GitCommit:"132a687512d7fb058d0f5890f07d4121b3f0a2e2", GitTreeState:"clean", BuildDate:"2021-05-12T12:32:49Z", GoVersion:"go1.15.12", Compiler:"gc", Platform:"linux/amd64"}

Bringing this conversation over from slack: https://redpandacommunity.slack.com/archives/C01AJDUT88N/p1643398053651669

It appears that librdkafka receives the wrong Kafka API version when it joins the group. This causes unexpected behavior when trying to use group instance IDs.

Minimal example to reproduce:

package main

import (
	"fmt"
	"github.com/confluentinc/confluent-kafka-go/kafka"
)

func main() {

	// this works and consumes messages as expected
	//c, err := kafka.NewConsumer(&kafka.ConfigMap{
	//	"bootstrap.servers": "myHost",
	//	"group.id":          "myGroup",
	//	"auto.offset.reset": "earliest",
	//})

	// this does not work and seems to just hang
	c, err := kafka.NewConsumer(&kafka.ConfigMap{
		"bootstrap.servers": "myHost",
		"group.id":          "myGroup",
		"auto.offset.reset": "earliest",
		"group.instance.id": "myGroup/0",
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(kafka.LibraryVersion())

	c.SubscribeTopics([]string{"myTopic", "^aRegex.*[Tt]opic"}, nil)

	for {
		msg, err := c.ReadMessage(-1)
		if err == nil {
			fmt.Printf("Message on %s: %s\n", msg.TopicPartition, string(msg.Value))
			break
		} else {
			// The client will automatically try to recover from all errors.
			fmt.Printf("Consumer error: %v (%v)\n", err, msg)
		}
	}

	c.Close()
}

[chris-kimberley]

We've seen the same issue using the Python confluent-kafka library (also librdkafka based).

[dotnwat]

Hi @BalazsHoranyi and @chris-kimberley Redpanda doesn't currently support static group membership which is why you are seeing the errors. Disabling use of group instance id should help until we have support for this feature.

[emaxerrno]

ah boom. static assignment. makes sense.

[chris-kimberley]

Ah. Do we have an idea for when this would be supported? It's pretty important for our use-case.

[emaxerrno]

@vsaraswat - we should add this to the compat work

@chris-kimberley if you subscribe to this tix it will notify you when we add

[rkruze]

This is a duplicate of #1335

[emaxerrno]

landed in dev.