private bool Rebalance(Cluster cluster)
{
var myTopicThreadIdsMap =
TopicCount.ConstructTopicCount(group, consumerIdString, parent.zkClient)
.GetConsumerThreadIdsPerTopic();
var consumersPerTopicMap = ZkUtils.GetConsumersPerTopic(parent.zkClient, group);
var brokers = ZkUtils.GetAllBrokersInCluster(parent.zkClient);
if (brokers.Count == 0)
{
// This can happen in a rare case when there are no brokers available in the cluster when the consumer is started.
// We log an warning and register for child changes on brokers/id so that rebalance can be triggered when the brokers
// are up.
Logger.Warn("no brokers found when trying to rebalance.");
parent.zkClient.SubscribeChildChanges(ZkUtils.BrokerIdsPath, parent.loadBalancerListener);
return(true);
}
else
{
var partitionsAssignmentPerTopicMap = ZkUtils.GetPartitionAssignmentForTopics(
parent.zkClient, myTopicThreadIdsMap.Keys.ToList());
var partitionsPerTopicMap = partitionsAssignmentPerTopicMap.ToDictionary(
p => p.Key, p => p.Value.Keys.OrderBy(x => x).ToList());
/**
* fetchers must be stopped to avoid Data duplication, since if the current
* rebalancing attempt fails, the partitions that are released could be owned by another consumer.
* But if we don't stop the fetchers first, this consumer would continue returning Data for released
* partitions in parallel. So, not stopping the fetchers leads to duplicate Data.
*/
this.CloseFetchers(cluster, (IDictionary <string, IList <KafkaStream <TKey, TValue> > >)KafkaMessageAndMetadataStreams, myTopicThreadIdsMap);
this.ReleasePartitionOwnership(parent.topicRegistry);
var partitionOwnershipDecision = new Dictionary <Tuple <string, int>, string>();
var currentTopicRegistry = new Pool <string, Pool <int, PartitionTopicInfo> >();
foreach (var topicAndConsumerThreadIsSet in myTopicThreadIdsMap)
{
var topic = topicAndConsumerThreadIsSet.Key;
var consumerThreadIdSet = topicAndConsumerThreadIsSet.Value;
currentTopicRegistry[topic] = new Pool <int, PartitionTopicInfo>();
var topicDirs = new ZKGroupTopicDirs(group, topic);
var curConsumers = consumersPerTopicMap.Get(topic);
var curPartitions = partitionsPerTopicMap.Get(topic);
var nPartsPerConsumer = curPartitions.Count / curConsumers.Count;
var nConsumersWithExtraPart = curPartitions.Count % curConsumers.Count;
Logger.InfoFormat("Consumer {0} rebalancing the following partitions: {1} for topic {2} with consumers: {3}", consumerIdString, string.Join(",", curPartitions), topic, string.Join(",", curConsumers));
foreach (var consumerThreadId in consumerThreadIdSet)
{
var myConsumerPosition = curConsumers.IndexOf(consumerThreadId);
Contract.Assert(myConsumerPosition >= 0);
var startPart = (nPartsPerConsumer * myConsumerPosition)
+ Math.Min(nConsumersWithExtraPart, myConsumerPosition);
var nParts = nPartsPerConsumer + (myConsumerPosition + 1 > nConsumersWithExtraPart ? 0 : 1);
/**
* Range-partition the sorted partitions to consumers for better locality.
* The first few consumers pick up an extra partition, if any.
*/
if (nParts <= 0)
{
Logger.WarnFormat(
"No broker partitions consumed by consumer thread {0} for topic {1}",
consumerThreadId,
topic);
}
else
{
for (var i = startPart; i < startPart + nParts; i++)
{
var partition = curPartitions[i];
Logger.InfoFormat("{0} attempting to claim partition {1}", consumerThreadId, partition);
this.AddPartitionTopicInfo(currentTopicRegistry, topicDirs, partition, topic, consumerThreadId);
// record the partition ownership decision
partitionOwnershipDecision[Tuple.Create(topic, partition)] = consumerThreadId;
}
}
}
}
/**
* move the partition ownership here, since that can be used to indicate a truly successful rebalancing attempt
* A rebalancing attempt is completed successfully only after the fetchers have been started correctly
*/
if (this.ReflectPartitionOwnershipDecision(partitionOwnershipDecision))
{
Logger.Info("Updating the cache");
Logger.Debug("Partitions per topic cache " + JObject.FromObject(partitionsPerTopicMap).ToString(Formatting.None));
Logger.Debug("Consumers per topic cache " + JObject.FromObject(consumersPerTopicMap).ToString(Formatting.None));
parent.topicRegistry = currentTopicRegistry;
this.UpdateFetcher(cluster);
return(true);
}
else
{
return(false);
}
}
}