Created on 02-03-2016 02:48 PM - edited 09-16-2022 03:01 AM
Attempting to provision a new environment containing cluster of 3 Kafka and Zookeeper nodes using the cloudera-director bootstrap-remote CLI utility. The bootstrap process fails with the following stack trace:
http://pastebin.com/Qyg9bL96
However, when a new cluster is created from the director UI, the process succeeds.
Here is our configuration file:
name: Atlas-Test
provider {
type: aws
publishAccessKeys: true
region: us-west-2
subnetId: subnet-daf4c9ad
securityGroupsIds: sg-feae6f99
instanceNamePrefix: test
associatePublicIpAddresses: false
}
ssh {
username: ec2-user
privateKey: /home/ec2-user/atlas.pem
}
instances {
default {
type: m4.large
image: ami-414b7271
tags {
owner: ${?USER}
}
}
}
cloudera-manager {
instance: ${instances.default} {
tags {
application: "Cloudera Manager 5"
}
}
enableEnterpriseTrial: true
}
cluster {
products {
CDH: 5
}
services: [ZOOKEEPER, KAFKA]
master {
count: 3
instance: ${instances.default} {
tags {
group: master
}
}
roles {
ZOOKEEPER: [SERVER]
}
}
gateway {
count: 3
instance: ${instances.default} {
tags {
group: kafka
}
}
roles {
KAFKA: [KAFKA_BROKER]
}
}
}
Created 02-04-2016 04:13 PM
The underlying cause of the failure seems to be a failure to start Kafla:
[2016-02-01 19:10:58] ERROR [pipeline-thread-35] - c.c.l.b.UnboundedWaitForApiCommand: Command Start with ID 72 failed. Details: ApiCommand{id=72, name=Start, startTime=Mon Feb 01 19:10:10 EST 2016, endTime=Mon Feb 01 19:10:10 EST 2016, active=false, success=false, resultMessage=Failed to start service., serviceRef=ApiServiceRef{peerName=null, clusterName=Atlas-Test, serviceName=CD-KAFKA-MnViYSKk}, roleRef=null, hostRef=null, parent=null}
Cloudera Manager command execution history should have more details.
I believe get this error because the Kafka parcel is not distributed (as you know Kafka is not included in the CDH parcel). Please add "KAFKA: 0.8.2" to the products {} block. You may also need to add reference to the precise parcel repositories you want to use.
Created 02-04-2016 04:13 PM
The underlying cause of the failure seems to be a failure to start Kafla:
[2016-02-01 19:10:58] ERROR [pipeline-thread-35] - c.c.l.b.UnboundedWaitForApiCommand: Command Start with ID 72 failed. Details: ApiCommand{id=72, name=Start, startTime=Mon Feb 01 19:10:10 EST 2016, endTime=Mon Feb 01 19:10:10 EST 2016, active=false, success=false, resultMessage=Failed to start service., serviceRef=ApiServiceRef{peerName=null, clusterName=Atlas-Test, serviceName=CD-KAFKA-MnViYSKk}, roleRef=null, hostRef=null, parent=null}
Cloudera Manager command execution history should have more details.
I believe get this error because the Kafka parcel is not distributed (as you know Kafka is not included in the CDH parcel). Please add "KAFKA: 0.8.2" to the products {} block. You may also need to add reference to the precise parcel repositories you want to use.