Support Questions
Find answers, ask questions, and share your expertise
Announcements
Alert: Welcome to the Unified Cloudera Community. Former HCC members be sure to read and learn how to activate your account here.

ERROR: ClusterDock installation

ERROR: ClusterDock installation

New Contributor

Hello,

 

I tried to setup and spin up the CDH cluster on my local machine running on Ubuntu 16.04 LTS, but getting below error and it doesnt start the cluster.

 

command used: $ clusterdock_run ./bin/start_cluster cdh

latest: Pulling from cloudera/clusterdock
Digest: sha256:8b125ab1c3c1edfaac9d6fd628586580bfcfa41cf72c5795151ed8e94c080bb2
Status: Image is up to date for cloudera/clusterdock:latest

.

.

.

.

.

 

Traceback (most recent call last):
File "./bin/start_cluster", line 70, in <module>
main()
File "./bin/start_cluster", line 63, in main
actions.start(args)
File "/root/clusterdock/clusterdock/topologies/cdh/actions.py", line 146, in start
raise Exception('Failed to start cluster.')
Exception: Failed to start cluster.

24 REPLIES 24

Re: ERROR: ClusterDock installation

Rising Star

Hi cloudraj,

 

How much RAM/how many CPU cores does your machine have? Is the error reproducible or intermittent?

 

-Dima

 

 

Re: ERROR: ClusterDock installation

New Contributor

Hi Dima,

 

I have 16GB RAM and I guess it have 4CPU cores.

 

This error is reproducible, Note: this is first time I am trying to run the cluster, i tried 3 times and everytime getting the same error.

 

Anyway thanks for your quick response.

Re: ERROR: ClusterDock installation

Rising Star
Hm, so the error you're seeing is that Cloudera Manager returns an error
while starting the cluster. Can you go into the CM web UI (its URL is
listed during start) and look at the recently-run commands and see where
it's failing? Also, can you monitor how much RAM is free on your machine
when start_cluster is executed? We typically want 16 GB of RAM free just
for the clusterdock process, not in total on the machine, so it might just
be a lack of system resources.

-Dima



--
-Dima

Re: ERROR: ClusterDock installation

New Contributor

I have the same issue using the latest Docker on Mac OS, and the latest clusterdock version 

cdh580_cm581* . My macbook has 4 cores and 16GB of RAM. The UI URL provided in the console output doesn't respond.

 

INFO:clusterdock.topologies.cdh.actions:CM server is now accessible at http://moby:32768

INFO:clusterdock.topologies.cdh.cm:Detected CM API v13.

INFO:clusterdock.topologies.cdh.cm_utils:Updating database configurations...

INFO:clusterdock.topologies.cdh.cm:Updating NameNode references in Hive metastore...

INFO:clusterdock.topologies.cdh.actions:Deploying client configuration...

INFO:clusterdock.topologies.cdh.actions:Starting cluster...

Traceback (most recent call last):

  File "./bin/start_cluster", line 70, in <module>

    main()

  File "./bin/start_cluster", line 63, in main

    actions.start(args)

  File "/root/clusterdock/clusterdock/topologies/cdh/actions.py", line 146, in start

    raise Exception('Failed to start cluster.')

Exception: Failed to start cluster.

 

Re: ERROR: ClusterDock installation

Rising Star
If using Docker for Mac (unsupported), you'd need to access "localhost" in place of the hostname displayed above.

Seeing as how you're running Docker for Mac, how much RAM and CPU are you allocating to the Docker instance itself? I'm guessing you're just starved on resources.

Re: ERROR: ClusterDock installation

New Contributor

I am using Docker native app for OS X (it doesn't use Virtual box anymore). I set it to allow Docker to use 8GB of RAM, and I belive Docker containers should be able to use all of 8GB, the docker inspect on the container says "Memory":0 which I think means use all available. Does it nead more than 8GB in the default mode?

 

I do see the login screen when I go to http://localhost:32768/

 

Re: ERROR: ClusterDock installation

Rising Star

Docker for Mac still runs in a VM (unless you figured out how to get OS X to use Linux :-p). But yes, as described on our blog post and in the Docker Hub description, you should have about 16 GB available to start up a 2-node cluster with all included services.

Re: ERROR: ClusterDock installation

New Contributor

Hi,

 

great job. I was already looking for a setup to run a small cluster on my laptop. 

However, I'm having the same issues on my Mac Book Pro with 4 CPU Cores and 16 GB using Docker for Mac.

 

Following the advice regarding the available memory, I configured Docker to use all 16 GB of memory and closed all other Apps. Thus, I finally managed to start the cluster with 'clusterdock_run ./bin/start_cluster cdh' without seeing any error messages.

 

But, some problems still remain:

  •  First, I had to figure out how to log into Cloudera Manager as this is not documented (yet). 'docker ps' shows that the primary node is exposing port 32768 (Seems like no port mapping is defined for CM). So, I managed to login at localhost:32768 using admin:admin as credentials.
  • CM shows that the Impala deamon and Lily HBase Indexer are not running because they had errors. The latter error log says "ZooKeeper session expired, shutting down."
  • Unfortunately, I'm not able to access Hue or any other hadoop service. It seems like the ports are not mapped in docker.

 

Hence, I would like to know

  • if the memory requirements be lowered? In the past, I could run CDH in VirtualBox with just 6-8 GB RAM.
  • how to access Hue, Hive, Spark through a Web browser without entering the container or digging into the docker configuration?

 

Thanks

 

 

 

Re: ERROR: ClusterDock installation

Rising Star

Hey protocluster,

 

So during cluster startup, it should have output the hostname:port that the CM server is being redirected to in the terminal (though with Docker for Mac, the hostname should be replaced with 'localhost' since there's some VM magic going on there). Good advice about making sure to publish admin:admin as the login creds; I took this for granted after being around this stuff for too long; I'll fix it.

 

Can't comment on the service startup failures; can't reproduce it on this end. Is it repeatable on your machine?

 

Yeah, I should add support for Hue port redirection. As for the others, that's an unavoidable reality of dealing with containers at the moment; they're using internally-accessible hostnames and ports that can't be accessed over the internet. No quick fix there, unfortunately. :(