Guodong Xu created BIGTOP-3205:
----------------------------------

             Summary: zookeeper servers didn't deploy in Clustered 
(Multi-Server) / ensemble mode
                 Key: BIGTOP-3205
                 URL: https://issues.apache.org/jira/browse/BIGTOP-3205
             Project: Bigtop
          Issue Type: Bug
          Components: deployment
    Affects Versions: 1.3.0
            Reporter: Guodong Xu


When I use './provisioner/docker/docker-hadoop.sh -C sample.yaml -c 3' to 
deploy a multi-node bigtop cluster, and include 'zookeeper' in sample.yaml's 
components list, such as:

 
{code:java}
components: [hdfs, yarn, mapreduce, zookeeper, kafka]{code}
 

By doing this, a zookeeper server will be launched on each of the container 
nodes, totally three of them.

Expected result is these three zookeeper servers should be connected to each 
other and form a zookeeper ensemble.

Ref: [https://zookeeper.apache.org/doc/r3.4.6/zookeeperAdmin.html]  Section: 
Clustered (Multi-Server) Setup.

However, the actual result is they are all standalone. And Kafka-servers only 
uses the zookeeper server in head node (node 1). Zookeeper servers on node 2 
and node 3 are wasted, nobody knows them.

More information: have a check of '/etc/zookeeper/conf/zoo.cfg', they all 
points to a single server which is the node 1's. Eg.

  server.0=bdd7eae82f27.bigtop.apache.org:2888:3888

Expected I should see three server.<id> lines in conf/zoo.cfg for a three-node 
zookeeper cluster.

 

The same error exists when deploying Bigtop onto multiple physical machines 
using 'puppet apply' and hiera.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to