对于使用合流python或enterpriseapi的用户。可以这样做：

def count_partitions(my_partitions) -> int:
    count = 0
    for part in my_partitions:
        count = count + 1
    return count

cluster_data: ClusterMetadata = producer.list_topics(topic=TOPIC)
    topic_data: TopicMetadata = cluster_data.topics[TOPIC]
    available_partitions: PartitionMetadata = topic_data.partitions
    print(count_partitions(available_partitions))

我在试图解决这个完全相同的问题时发现了这个问题。我知道这个问题由来已久，但下面是我提出的解决方案（使用kazoo与Zookeeper交谈）：

from kazoo.client import KazooClient

class KafkaInfo(object):
    def __init__(self, hosts):
        self.zk = KazooClient(hosts)
        self.zk.start()

    def topics(self):
        return self.zk.get_children('/brokers/topics')

    def partitions(self, topic):
        strs = self.zk.get_children('/brokers/topics/%s/partitions' % topic)
        return map(int, strs)

    def consumers(self):
        return self.zk.get_children('/consumers')

    def topics_for_consumer(self, consumer):
        return self.zk.get_children('/consumers/%s/offsets' % consumer)

    def offset(self, topic, consumer, partition):
        (n, _) = self.zk.get('/consumers/%s/offsets/%s/%d' % (consumer, topic, partition))
        return int(n)

可能是一个稍微简单化的解决方案，但是：

from kafka import KafkaClient

client = KafkaClient('SERVER:PORT')
topic_partition_ids = client.get_partition_ids_for_topic(b'TOPIC')
len(topic_partition_ids)

在python3.4.3/kafka-python0.9.3上测试