Help: kafka.admin.AdminOperationException: replication factor: 3 larger than available brokers: 1

[fillest]

Hello.
Yesterday I got my kafka broker killed by OOM killer and after its start I'm seeing series of strange errors in the log:

[2016-07-31 23:50:47,402] INFO [Kafka Server 0], started (kafka.server.KafkaServer)
[2016-07-31 23:50:48,022] ERROR [KafkaApi-0] Error when handling request {group_id=kafka-python-default-group} (kafka.server.KafkaApis)
kafka.admin.AdminOperationException: replication factor: 3 larger than available brokers: 1
        at kafka.admin.AdminUtils$.assignReplicasToBrokers(AdminUtils.scala:117)
        at kafka.admin.AdminUtils$.createTopic(AdminUtils.scala:403)
        at kafka.server.KafkaApis.kafka$server$KafkaApis$$createTopic(KafkaApis.scala:629)
        at kafka.server.KafkaApis.kafka$server$KafkaApis$$createGroupMetadataTopic(KafkaApis.scala:651)
        at kafka.server.KafkaApis$$anonfun$getOrCreateGroupMetadataTopic$1.apply(KafkaApis.scala:657)
        at kafka.server.KafkaApis$$anonfun$getOrCreateGroupMetadataTopic$1.apply(KafkaApis.scala:657)
        at scala.Option.getOrElse(Option.scala:121)
        at kafka.server.KafkaApis.getOrCreateGroupMetadataTopic(KafkaApis.scala:657)
        at kafka.server.KafkaApis.handleGroupCoordinatorRequest(KafkaApis.scala:818)
        at kafka.server.KafkaApis.handle(KafkaApis.scala:86)
        at kafka.server.KafkaRequestHandler.run(KafkaRequestHandler.scala:60)
        at java.lang.Thread.run(Thread.java:745)

"group_id=kafka-python-default-group" looks like a strong indication that the request came from python client. Why could this possibly happen at all and why does it use replication factor 3?
I'm using kafka-python==1.2.1, a single kafka 0.10 broker with replication factor 1 with a single local producer and a single remote (different host) consumer. Automatic topic creation is disabled on the broker. All topics were created in advance during a deploy long ago.

[brynmathias]

This often happens to me when the kafka server has the wrong broker number in zookeeper, which can happen when kafka doesn't shut down gracefully and the kafka instance that you bring back up has a broker number that already exists.

I think I have fixed it before with:
shut down kafka broker.
Shut down zookeeper,
Start zookeeperm
Start kafka.

[alexeybaj]

Value of 3 is hardcoded in OffsetManager.scala

val DefaultOffsetsTopicReplicationFactor = 3.toShort

and is applied during special offsets topic creation if number of alive brokers is zero. So it is very likely cluster issue.

[dpkp]

This is definitely a broker configuration issue, unrelated to kafka-python. The topic that is failing to create is __consumer_offsets . Try setting offsets.topic.replication.factor to 1 on your single broker.