I have a CDH5.2 installation with yarn, hdfs, spark 1.1.0. I was able to successfully install CM5 using the manual path B:
Now, I am able to see the cloudera manager admin screen, and also able to see my nodes under "Hosts" tab. when you go "Hosts>> Components" I am able to see my installed components. But not able to see anything under the "Cluster" tab. How do I go about configuring the cluster and the various services.
Please let me know if I am on the right path. how do i move forward now?
As related to your post in one of the other CM discussions (thanks for creating a unique post, we'll remove that other one).
Review the discussion on the thread summary list (link below)
If you do not have significant data or can migrate it off your existing CDH, I would suggest reviewing any non-default settings you made, and effectively start over by installing CM first, then using parcel based deployment to deploy CDH, and import your data back in. There is not an automated method for ingesting an existing cluster into CM. Installing CM over an existing cluster can be a complicated and manual process, there is discussion in these threads on the subject:
Thanks for responding so quickly.
I do not have significant data on my cluster, so I can go with this approach. I had installed the cdh components by adding the CDH 5 repository and then using apt-get install.. I followed most of the default configuration.
As I understand the process:
1) Back up my configuration files, or any other non default configurations.
2) Install the components again via CM5.
3) Define services/roles.
4) reapply the configurations from step 1.
Do I need to un install the earlier cdh packages before doing the above steps?
And in step 2, will CM install the existing components again?
Thanks for the support!
There are 2 approaches to running CDH in a CM cluster, using packages (the original way) and Parcel ( the better way). With parcels you are able to do things like rolling upgrades, rolling restarts, etc.
Removing the packages is discussed here
This link has good information on cleaning up default paths as well:
Consider file cleanup is an important step, the CM setup wizard will not destroy existing data if it finds it in the default paths (it will stop and fail at that point). As you stated no signfiicant data is in there so its OK to remove your old HDFS base path and child paths as indicated in that second link.
Remove the packages before attempting to deploy and configure parcels for CDH within CM. The important thing to save from your old install is your xml configuration files (if you had changed settings for specific needs of your config). Otherwise all the defaults will pretty much be re-created as CM creates the cluster.
The wizard in CM will allow you to add the cluster hosts, and then specify what CDH services you want to run across the servers and then go through and set things up.
1) I have to remove cdh components like hadoop-namenode, etc.. (which I installed using repository, then apt-get install)
2) Remove all the directories I created for different hadoop componenets like /data/1/dfs/nn etc..
basically doing ' hadoop fs -rmr / ' and also removing the local directories.
3) Start the CM admin console and run the CM wizard to install cdh components using Parcels. (which will download everything again and install)
Cant we make use of the existing installation of cdh components? May be the CM wizard finds that oh... the components are alraedy installed so it moves to the next step of defining roles and services? Do we really need to remove them in step 1?
Thanks for the support!