samza-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Davide Simoncelli <netcelli....@gmail.com>
Subject Containers stuck in event loop
Date Tue, 02 Jun 2015 07:44:15 GMT
Hello,

I have had problems running my Samza application on the cluster. The application starts fine
and so the main event loop. As soon I start to send messages to Kafka, Samza doesn’t start
the Kafka system consumer (there are no logs that state that).
The CPU usage for all containers is about 100% even if I stop producers. It is like the container
is stuck and can’t start the consumer. However Samza can set the offsets for different partitions.
For example in a container I see:

o.a.samza.system.kafka.GetOffset - Able to successfully read from offset 0 for topic and partition
[test_topic,40]. Using it to instantiate consumer.
o.a.samza.system.kafka.BrokerProxy - Starting BrokerProxy for node1.cluster.com:9092
o.a.samza.system.kafka.GetOffset - Validating offset 0 for topic and partition [test_topic,19]
o.a.samza.system.kafka.GetOffset - Able to successfully read from offset 0 for topic and partition
[test_topic,19]. Using it to instantiate consumer.
o.a.samza.container.SamzaContainer - Entering run loop.

Here is the configuration I use:
{
yarn.container.count=24,
 systems.kafka.samza.key.serde=int,
 systems.kafka.consumer.zookeeper.connect=localhost:2181/,
 serializers.registry.int.class=org.apache.samza.serializers.StringSerdeFactory,
 systems.kafka.samza.factory=org.apache.samza.system.kafka.KafkaSystemFactory,
 task.drop.deserialization.errors=true,
 yarn.container.memory.mb=1024,
 task.inputs=kafka.test_topic,
 job.factory.class=org.apache.samza.job.yarn.YarnJobFactory,
 yarn.package.path=hdfs://node1.cluster.com:8020/my-app-0.0.1-dist.tar.gz,
 task.class=com.company.test.Task,
 systems.kafka.samza.msg.serde=json,
 job.name=test,
 serializers.registry.json.class=org.apache.samza.serializers.JsonSerdeFactory,
 systems.kafka.producer.bootstrap.servers=node1.cluster.com:9092,node2.cluster.com:9092,node3.cluster.com:9092,
 }

The application was working before a server crash. I tried to clean all Zookeeper data and
restart everything. Do you have any idea why the consumer doesn’t work?

Regards

Davide
Mime
View raw message