Finish breaking out integration tests

Author: wizzat

test/test_client_integration.py

@@ -0,0 +1,47 @@
+import unittest
+import time
+
+from kafka import *  # noqa
+from kafka.common import *  # noqa
+from kafka.codec import has_gzip, has_snappy
+from .fixtures import ZookeeperFixture, KafkaFixture
+from .testutil import *
+
+@unittest.skipIf(skip_integration(), 'Skipping Integration')
+class TestKafkaClientIntegration(KafkaIntegrationTestCase):
+    @classmethod
+    def setUpClass(cls):  # noqa
+        cls.zk = ZookeeperFixture.instance()
+        cls.server = KafkaFixture.instance(0, cls.zk.host, cls.zk.port)
+
+    @classmethod
+    def tearDownClass(cls):  # noqa
+        cls.server.close()
+        cls.zk.close()
+
+    def test_consume_none(self):
+        fetch = FetchRequest(self.topic, 0, 0, 1024)
+
+        fetch_resp, = self.client.send_fetch_request([fetch])
+        self.assertEquals(fetch_resp.error, 0)
+        self.assertEquals(fetch_resp.topic, self.topic)
+        self.assertEquals(fetch_resp.partition, 0)
+
+        messages = list(fetch_resp.messages)
+        self.assertEquals(len(messages), 0)
+
+    ####################
+    #   Offset Tests   #
+    ####################
+
+    @unittest.skip('commit offset not supported in this version')
+    def test_commit_fetch_offsets(self):
+        req = OffsetCommitRequest(self.topic, 0, 42, "metadata")
+        (resp,) = self.client.send_offset_commit_request("group", [req])
+        self.assertEquals(resp.error, 0)
+
+        req = OffsetFetchRequest(self.topic, 0)
+        (resp,) = self.client.send_offset_fetch_request("group", [req])
+        self.assertEquals(resp.error, 0)
+        self.assertEquals(resp.offset, 42)
+        self.assertEquals(resp.metadata, "")  # Metadata isn't stored for now

test/test_consumer_integration.py

@@ -0,0 +1,208 @@
+import unittest
+from datetime import datetime
+
+from kafka import *  # noqa
+from kafka.common import *  # noqa
+from kafka.consumer import MAX_FETCH_BUFFER_SIZE_BYTES
+from .fixtures import ZookeeperFixture, KafkaFixture
+from .testutil import *
+
+@unittest.skipIf(skip_integration(), 'Skipping Integration')
+class TestConsumerIntegration(KafkaIntegrationTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.zk = ZookeeperFixture.instance()
+        cls.server1 = KafkaFixture.instance(0, cls.zk.host, cls.zk.port)
+        cls.server2 = KafkaFixture.instance(1, cls.zk.host, cls.zk.port)
+
+        cls.server = cls.server1 # Bootstrapping server
+
+    @classmethod
+    def tearDownClass(cls):  # noqa
+        cls.server1.close()
+        cls.server2.close()
+        cls.zk.close()
+
+    def send_messages(self, partition, messages):
+        messages = [ create_message(self.msg(str(msg))) for msg in messages ]
+        produce = ProduceRequest(self.topic, partition, messages = messages)
+        resp, = self.client.send_produce_request([produce])
+        self.assertEquals(resp.error, 0)
+
+        return [ x.value for x in messages ]
+
+    def assert_message_count(self, messages, num_messages):
+        # Make sure we got them all
+        self.assertEquals(len(messages), num_messages)
+
+        # Make sure there are no duplicates
+        self.assertEquals(len(set(messages)), num_messages)
+
+    def test_simple_consumer(self):
+        self.send_messages(0, range(0, 100))
+        self.send_messages(1, range(100, 200))
+
+        # Start a consumer
+        consumer = SimpleConsumer(self.client, "group1",
+                                  self.topic, auto_commit=False,
+                                  iter_timeout=0)
+
+        self.assert_message_count([ message for message in consumer ], 200)
+
+        consumer.stop()
+
+    def test_simple_consumer__seek(self):
+        self.send_messages(0, range(0, 100))
+        self.send_messages(1, range(100, 200))
+
+        consumer = SimpleConsumer(self.client, "group1",
+                                  self.topic, auto_commit=False,
+                                  iter_timeout=0)
+
+        # Rewind 10 messages from the end
+        consumer.seek(-10, 2)
+        self.assert_message_count([ message for message in consumer ], 10)
+
+        # Rewind 13 messages from the end
+        consumer.seek(-13, 2)
+        self.assert_message_count([ message for message in consumer ], 13)
+
+        consumer.stop()
+
+    def test_simple_consumer_blocking(self):
+        consumer = SimpleConsumer(self.client, "group1",
+                                  self.topic,
+                                  auto_commit=False, iter_timeout=0)
+
+        # Ask for 5 messages, nothing in queue, block 5 seconds
+        with Timer() as t:
+            messages = consumer.get_messages(block=True, timeout=5)
+            self.assert_message_count(messages, 0)
+        self.assertGreaterEqual(t.interval, 5)
+
+        self.send_messages(0, range(0, 10))
+
+        # Ask for 5 messages, 10 in queue. Get 5 back, no blocking
+        with Timer() as t:
+            messages = consumer.get_messages(count=5, block=True, timeout=5)
+            self.assert_message_count(messages, 5)
+        self.assertLessEqual(t.interval, 1)
+
+        # Ask for 10 messages, get 5 back, block 5 seconds
+        with Timer() as t:
+            messages = consumer.get_messages(count=10, block=True, timeout=5)
+            self.assert_message_count(messages, 5)
+        self.assertGreaterEqual(t.interval, 5)
+
+        consumer.stop()
+
+    def test_simple_consumer_pending(self):
+        # Produce 10 messages to partitions 0 and 1
+        self.send_messages(0, range(0, 10))
+        self.send_messages(1, range(10, 20))
+
+        consumer = SimpleConsumer(self.client, "group1", self.topic,
+                                  auto_commit=False, iter_timeout=0)
+
+        self.assertEquals(consumer.pending(), 20)
+        self.assertEquals(consumer.pending(partitions=[0]), 10)
+        self.assertEquals(consumer.pending(partitions=[1]), 10)
+
+        consumer.stop()
+
+    def test_multi_process_consumer(self):
+        # Produce 100 messages to partitions 0 and 1
+        self.send_messages(0, range(0, 100))
+        self.send_messages(1, range(100, 200))
+
+        consumer = MultiProcessConsumer(self.client, "grp1", self.topic, auto_commit=False)
+
+        self.assert_message_count([ message for message in consumer ], 200)
+
+        consumer.stop()
+
+    def test_multi_process_consumer_blocking(self):
+        consumer = MultiProcessConsumer(self.client, "grp1", self.topic, auto_commit=False)
+
+        # Ask for 5 messages, No messages in queue, block 5 seconds
+        with Timer() as t:
+            messages = consumer.get_messages(block=True, timeout=5)
+            self.assert_message_count(messages, 0)
+
+        self.assertGreaterEqual(t.interval, 5)
+
+        # Send 10 messages
+        self.send_messages(0, range(0, 10))
+
+        # Ask for 5 messages, 10 messages in queue, block 0 seconds
+        with Timer() as t:
+            messages = consumer.get_messages(count=5, block=True, timeout=5)
+            self.assert_message_count(messages, 5)
+        self.assertLessEqual(t.interval, 1)
+
+        # Ask for 10 messages, 5 in queue, block 5 seconds
+        with Timer() as t:
+            messages = consumer.get_messages(count=10, block=True, timeout=5)
+            self.assert_message_count(messages, 5)
+        self.assertGreaterEqual(t.interval, 5)
+
+        consumer.stop()
+
+    def test_multi_proc_pending(self):
+        self.send_messages(0, range(0, 10))
+        self.send_messages(1, range(10, 20))
+
+        consumer = MultiProcessConsumer(self.client, "group1", self.topic, auto_commit=False)
+
+        self.assertEquals(consumer.pending(), 20)
+        self.assertEquals(consumer.pending(partitions=[0]), 10)
+        self.assertEquals(consumer.pending(partitions=[1]), 10)
+
+        consumer.stop()
+
+    def test_large_messages(self):
+        # Produce 10 "normal" size messages
+        small_messages = self.send_messages(0, [ str(x) for x in range(10) ])
+
+        # Produce 10 messages that are large (bigger than default fetch size)
+        large_messages = self.send_messages(0, [ random_string(5000) for x in range(10) ])
+
+        # Consumer should still get all of them
+        consumer = SimpleConsumer(self.client, "group1", self.topic,
+                                  auto_commit=False, iter_timeout=0)
+
+        expected_messages = set(small_messages + large_messages)
+        actual_messages = set([ x.message.value for x in consumer ])
+        self.assertEqual(expected_messages, actual_messages)
+
+        consumer.stop()
+
+    def test_huge_messages(self):
+        huge_message, = self.send_messages(0, [
+            create_message(random_string(MAX_FETCH_BUFFER_SIZE_BYTES + 10)),
+        ])
+
+        # Create a consumer with the default buffer size
+        consumer = SimpleConsumer(self.client, "group1", self.topic,
+                                  auto_commit=False, iter_timeout=0)
+
+        # This consumer failes to get the message
+        with self.assertRaises(ConsumerFetchSizeTooSmall):
+            consumer.get_message(False, 0.1)
+
+        consumer.stop()
+
+        # Create a consumer with no fetch size limit
+        big_consumer = SimpleConsumer(self.client, "group1", self.topic,
+                                      max_buffer_size=None, partitions=[0],
+                                      auto_commit=False, iter_timeout=0)
+
+        # Seek to the last message
+        big_consumer.seek(-1, 2)
+
+        # Consume giant message successfully
+        message = big_consumer.get_message(block=False, timeout=10)
+        self.assertIsNotNone(message)
+        self.assertEquals(message.message.value, huge_message)
+
+        big_consumer.stop()

test/test_failover_integration.py

@@ -0,0 +1,116 @@
+import unittest
+import time
+
+from kafka import *  # noqa
+from kafka.common import *  # noqa
+from .fixtures import ZookeeperFixture, KafkaFixture
+from .testutil import *
+
+@unittest.skipIf(skip_integration(), 'Skipping Integration')
+class TestFailover(KafkaIntegrationTestCase):
+    create_client = False
+
+    @classmethod
+    def setUpClass(cls):  # noqa
+        zk_chroot = random_string(10)
+        replicas = 2
+        partitions = 2
+
+        # mini zookeeper, 2 kafka brokers
+        cls.zk = ZookeeperFixture.instance()
+        kk_args = [cls.zk.host, cls.zk.port, zk_chroot, replicas, partitions]
+        cls.brokers = [KafkaFixture.instance(i, *kk_args) for i in range(replicas)]
+
+        hosts = ['%s:%d' % (b.host, b.port) for b in cls.brokers]
+        cls.client = KafkaClient(hosts)
+
+    @classmethod
+    def tearDownClass(cls):
+        cls.client.close()
+        for broker in cls.brokers:
+            broker.close()
+        cls.zk.close()
+
+    def test_switch_leader(self):
+        key, topic, partition = random_string(5), self.topic, 0
+        producer = SimpleProducer(self.client)
+
+        for i in range(1, 4):
+
+            # XXX unfortunately, the conns dict needs to be warmed for this to work
+            # XXX unfortunately, for warming to work, we need at least as many partitions as brokers
+            self._send_random_messages(producer, self.topic, 10)
+
+            # kil leader for partition 0
+            broker = self._kill_leader(topic, partition)
+
+            # expect failure, reload meta data
+            with self.assertRaises(FailedPayloadsError):
+                producer.send_messages(self.topic, 'part 1')
+                producer.send_messages(self.topic, 'part 2')
+            time.sleep(1)
+
+            # send to new leader
+            self._send_random_messages(producer, self.topic, 10)
+
+            broker.open()
+            time.sleep(3)
+
+            # count number of messages
+            count = self._count_messages('test_switch_leader group %s' % i, topic)
+            self.assertIn(count, range(20 * i, 22 * i + 1))
+
+        producer.stop()
+
+    def test_switch_leader_async(self):
+        key, topic, partition = random_string(5), self.topic, 0
+        producer = SimpleProducer(self.client, async=True)
+
+        for i in range(1, 4):
+
+            self._send_random_messages(producer, self.topic, 10)
+
+            # kil leader for partition 0
+            broker = self._kill_leader(topic, partition)
+
+            # expect failure, reload meta data
+            producer.send_messages(self.topic, 'part 1')
+            producer.send_messages(self.topic, 'part 2')
+            time.sleep(1)
+
+            # send to new leader
+            self._send_random_messages(producer, self.topic, 10)
+
+            broker.open()
+            time.sleep(3)
+
+            # count number of messages
+            count = self._count_messages('test_switch_leader_async group %s' % i, topic)
+            self.assertIn(count, range(20 * i, 22 * i + 1))
+
+        producer.stop()
+
+    def _send_random_messages(self, producer, topic, n):
+        for j in range(n):
+            resp = producer.send_messages(topic, random_string(10))
+            if len(resp) > 0:
+                self.assertEquals(resp[0].error, 0)
+        time.sleep(1)  # give it some time
+
+    def _kill_leader(self, topic, partition):
+        leader = self.client.topics_to_brokers[TopicAndPartition(topic, partition)]
+        broker = self.brokers[leader.nodeId]
+        broker.close()
+        time.sleep(1)  # give it some time
+        return broker
+
+    def _count_messages(self, group, topic):
+        hosts = '%s:%d' % (self.brokers[0].host, self.brokers[0].port)
+        client = KafkaClient(hosts)
+        consumer = SimpleConsumer(client, group, topic, auto_commit=False, iter_timeout=0)
+        all_messages = []
+        for message in consumer:
+            all_messages.append(message)
+        consumer.stop()
+        client.close()
+        return len(all_messages)