Docker Quickstart Image Issue: services in bad health

[#1]

I noticed that whenever I install the docker quickstart, run cloudera manager on it and then start the cluster, I see that the following services are red:

1) Host: 

Clock Offset

2) HBase:

Bad : Master summary: quickstart.cloudera (Availability: Active, Health: Bad). This health test reflects the health of the active Master.

Bad : This RegionServer is not connected to its cluster.

Bad : This role's process exited. This role is supposed to be started.

3) HDFS:

Bad : NameNode summary: quickstart.cloudera (Availability: Active, Health: Bad). This health test reflects the health of the active NameNode.

4) YARN:

Bad : ResourceManager summary: quickstart.cloudera (Availability: Active, Health: Bad). This health test reflects the health of the active ResourceManager.

My virtual machine is running Ubuntu 14.04, has 24 GB memory, 8 cores, 250 GB Storage. I am using the latest docker cloudera image using 

docker pull cloudera/quickstart:latest

Also, I am installing JDK 1.8 before I run 

/home/cloudera/cloudera-manager --enterprise

To install the JDK: I set the JAVA_HOME variable to the new path, upgrade /etc/profile to reflect the change and also add it to /etc/default/bigtop-utils. Once the Cloudera Manager is up, I update the Hosts Configuration with this change as well.

Currently the latest docker image has CDH 5.7.

What can I do to have all services in green upon running Cloudera quickstart?

[#2 · Accepted solution]

Hey Rohit,

Wanna give clusterdock a shot? It's described in a blog post [1] and is a bit more conventional when it comes to modeling what a cluster deployment looks like (which makes debugging more straightforward). It also runs the newer Cloudera 5.8 release.

1. http://blog.cloudera.com/blog/2016/08/multi-node-clusters-with-cloudera-quickstart-for-docker/

[#3]

Thanks for the info! Are there detailed steps for installation? I tried it, and it installs a container, which if I run starts a python shell. It would be good to have some detailed installation documentation on this.

[#4]

The blog post has step-by-step directions. You don't interact directly with the container but with the clusterdock_run commands because of the nature of what the framework does.

[#5]

So is there a way to use JDK 1.8 with the Clusterdock setup?

[#6]

We use the JDK version used by the Cloudera one-click install, but it's not easily configurable at start time.

[#7]

So does that mean I cannot change the JDK version (similar to a full install) over here? 

[#8]

Thanks @dspivak. One last question: I see that clusterdock created 2 nodes. Are these virtual nodes? 

My cluster itself has 2 nodes, each has 24 GB memory and 8 Cores. So:

1) How do I know if each of my physical node is being used as a clusterdock node.

2) Is there a way to map them?

3) Is there a way to create more than 2 nodes using clusterdock?

Again, thanks for your help!