DCOS Kafka Deployment Stuck on Vagrant Cluster

0
votes

I am using the latest DCOS release (0.8.0) and spun up a local vagrant cluster with 4 private agent nodes, one public agent node, one master and one boot machine 

vagrant up m1 a1 a2 a3 a4 p1 boot

I then installed kafka with the default configurations using the following command:

dcos package install kafka

After examining the marathon logs, it looks like mesos is declining the resource offers for kafka and is therefore not able to deploy all three brokers. I suspect that the issue is with a mismatch between the role of the resource offer and the desired kafka role.

Aggregate cluster statistics: CPU: 18 shares Memory: 7GiB

Here is a snapshot of the marathon logs:

s.offer.OfferEvaluator: Satisfying resource requirement: name: "cpus" type: SCALAR scalar { value: 1.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "" } } }
with resource: name: "cpus" type: SCALAR scalar { value: 1.0 } role: "*"
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Reserves Resource
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Fulfilled resource: name: "cpus" type: SCALAR scalar { value: 1.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "f8468a5a-8166-4d5d-a594-f197a8b356cd" } } }
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.MesosResourcePool: Retrieving resource for reservation
WARN  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Failed to satisfy resource requirement: name: "mem" type: SCALAR scalar { value: 2304.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "" } } }
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.MesosResourcePool: Retrieving resource for reservation
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Satisfying resource requirement: name: "cpus" type: SCALAR scalar { value: 0.5 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "" } } }
with resource: name: "cpus" type: SCALAR scalar { value: 0.5 } role: "*"
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Reserves Resource
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Fulfilled resource: name: "cpus" type: SCALAR scalar { value: 0.5 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "37ba47a3-7c6e-47a2-972c-d35d130ae8df" } } }
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.MesosResourcePool: Retrieving resource for reservation
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Satisfying resource requirement: name: "mem" type: SCALAR scalar { value: 256.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "" } } }
with resource: name: "mem" type: SCALAR scalar { value: 256.0 } role: "*"
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Reserves Resource
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Fulfilled resource: name: "mem" type: SCALAR scalar { value: 256.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "8f937e43-c80a-4434-a858-08e6bdb02cc2" } } }
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.MesosResourcePool: Retrieving resource for reservation
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Satisfying resource requirement: name: "cpus" type: SCALAR scalar { value: 1.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "" } } }
with resource: name: "cpus" type: SCALAR scalar { value: 1.0 } role: "*"
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Reserves Resource
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Fulfilled resource: name: "cpus" type: SCALAR scalar { value: 1.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "b568cfaa-c216-4479-9853-5bc4e6ee2d9e" } } }
INFO  [2016-08-02 16:28:08,919] org.apache.mesos.offer.MesosResourcePool: Retrieving resource for reservation
WARN  [2016-08-02 16:28:08,919] org.apache.mesos.offer.OfferEvaluator: Failed to satisfy resource requirement: name: "mem" type: SCALAR scalar { value: 2304.0 } role: "kafka-role" reservation { principal: "kafka-principal" labels { labels { key: "resource_id" value: "" } } }
WARN  [2016-08-02 16:28:08,919] org.apache.mesos.scheduler.plan.DefaultStageScheduler: Unable to find any offers which fulfill requirement provided by block broker-0: org.apache.mesos.offer.OfferRequirement@2a28d9e[avoidAgents=[],colocateAgents=[],taskRequirements=[org.apache.mesos.offer.TaskRequirement@5c8e5b0f],executorRequirement=org.apache.mesos.offer.ExecutorRequirement@6179902f[executorInfo=executor_id {
  value: ""
}
resources {
  name: "cpus"
  type: SCALAR
  scalar {
    value: 0.5
  }
  role: "kafka-role"
  reservation {
    principal: "kafka-principal"
    labels {
      labels {
        key: "resource_id"
        value: ""
      }
    }
  }
}
resources {
  name: "mem"
  type: SCALAR
  scalar {
    value: 256.0
  }
  role: "kafka-role"
  reservation {
    principal: "kafka-principal"
    labels {
      labels {
        key: "resource_id"
        value: ""
      }
    }
  }
}
command {
  uris {
    value: "https://downloads.mesosphere.com/kafka/assets/jre-8u91-linux-x64.tar.gz"
  }
  uris {
    value: "https://downloads.mesosphere.com/kafka/assets/kafka_2.11-0.10.0.0.tgz"
  }
  uris {
    value: "https://downloads.mesosphere.com/kafka/assets/1.1.9-0.10.0.0/overrider.zip"
  }
  uris {
    value: "https://downloads.mesosphere.com/kafka/assets/1.1.9-0.10.0.0/executor.zip"
  }
  environment {
    variables {
      name: "JAVA_HOME"
      value: "jre1.8.0_91"
    }
  }
  value: "./executor/bin/kafka-executor -Dlogback.configurationFile=executor/conf/logback.xml"
}
framework_id {
  value: "2601d122-2b59-4823-83f8-84351376d53c-0001"
}
name: "broker-0"
,resourceRequirements=[org.apache.mesos.offer.ResourceRequirement@39ba5849, org.apache.mesos.offer.ResourceRequirement@2f62ec9d]]]
INFO  [2016-08-02 16:28:08,919] com.mesosphere.dcos.kafka.plan.KafkaUpdateBlock: broker-0: changed status from: Pending to: Pending
INFO  [2016-08-02 16:28:08,920] com.mesosphere.dcos.kafka.scheduler.KafkaRepairScheduler: Missing brokerIds: []
INFO  [2016-08-02 16:28:08,921] org.apache.mesos.offer.ResourceCleaner: 4 offers with 16 resources => 0 destroy and 0 unreserve operations
INFO  [2016-08-02 16:28:08,921] com.mesosphere.dcos.kafka.scheduler.KafkaScheduler: Accepted 0 of 4 offers: []
INFO  [2016-08-02 16:28:08,921] com.mesosphere.dcos.kafka.scheduler.KafkaScheduler: Declining offer: 2601d122-2b59-4823-83f8-84351376d53c-O1343
INFO  [2016-08-02 16:28:08,921] com.mesosphere.dcos.kafka.scheduler.KafkaScheduler: Declining offer: 2601d122-2b59-4823-83f8-84351376d53c-O1344
INFO  [2016-08-02 16:28:08,921] com.mesosphere.dcos.kafka.scheduler.KafkaScheduler: Declining offer: 2601d122-2b59-4823-83f8-84351376d53c-O1345
INFO  [2016-08-02 16:28:08,921] com.mesosphere.dcos.kafka.scheduler.KafkaScheduler: Declining offer: 2601d122-2b59-4823-83f8-84351376d53c-O1346
INFO  [2016-08-02 16:28:14,928]
2
vagrantapache-kafkamesosdcos

2 Answers

0
votes

The Kafka framework in DCOS 1.8 a stateful service. It is specifically designed to allocate storage for each broker and maintain a link between the broker instance and that storage (so broker-0 is not re-deployed to a host that does not have the the storage initially allocated for broker-0). My experience is that the messages around the resource offers are not very clear when that storage constraint is not met. By default, the package will request 5 GB of dedicated storage for each broker.

I don't have a lot of experience with vagrant, but it could be that vagrant is not exposing the necessary "storage availability" for DC/OS to be able to honor the broker resource request.

0
votes

The problem was indeed as the logs reported: not enough memory to schedule kafka using the default settings. I was able to deploy kafka on DCOS vagrant with the default vagrant config and three agent nodes using the following kafka config: 

{
    "brokers": {
        "count": 1,
        "mem": 512,
        "disk": 1000
    }
}