06-10-2014 11:37 AM
I think there is discrepancy/confusion in documentation of CM 5.0.1 installation process
1. http://cloudera.com/content/cloudera-content/cloudera-docs/CM5/latest/Cloudera-Manager-Installation-... - asks to create web dir
2.http://cloudera.com/content/cloudera-content/cloudera-docs/CM5/latest/Cloudera-Manager-Managing-Clus... mentions /opt/cloudera/parcel-repo as local location on server
3.http://cloudera.com/content/cloudera-content/cloudera-docs/CM5/latest/Cloudera-Manager-Installation-... mentions parcel directory /opt/cloudera/parcel-repo
I am installing CM 5.0.1 on ubuntu vms using path C Tarball, I have created /opt/cloudera/parcel-repo and put parcel manifest and sha files there but I dont see the parcel in admin interface when adding cluster
also when adding hosts to cluster it tries to download oracle jdk 1.6 and doesn't find package and fails though I have jdk 1.7 on all hosts?
06-10-2014 03:05 PM
I imagine most of the confusion comes from the title, "Creating and Using a Local Parcel Repository". What that documentation describes is really about building a web server of your own to serve as the "remote parcel repository." The title, indeed, could be altered to be more accurate.
The http server you set up would be entered in the Cloudera Manager Configuration > Settings > Parcels > Remote Parcel Repository URLs. By default, they all point to archive.cloudera.com.
The "local" parcel repository (Local Parcel Repository Path in the CM Configuration Settings) points to the location on disk where the parcel and sha files are placed after downloading from the Remote Parcel Repository.
So, the expected way this works by default is that Cloudera Manager is configured to check one or more web servers every hour to find out if there are any new pacels there. You can configure CM to automatically download any new Parcels, but that is disabled by default.
When a parcel is downloaded (automatically or manually via the CM GUI), Cloudera Manager places the .parcel file and the generated .sha file in the Local Parcel Repository Path. It is OK to bypass the http server, as you have done, and manually place the parcel file and the sha file in the local parcel repository path.
The Parcel can then be distributed to other servers who are managed by Cloudera Manager.
Part 2 of your query focuses on why parecles are not showing up in the GUI. If you downloaded the parcel and created the .sha file and then placed them in the default Local Parcel Repository Path (/opt/cloudera/parcel-repo), but they do not show up, there are two things I suggest checking:
(1) File permissions
Make sure you followed the documentation and did the following:
mkdir -p /opt/cloudera/parcel-repo
chown username:groupname /opt/cloudera/parcel-repo
(2) If the server was running when you moved the parcel and sha files in, the server will not check what is in that directory until the next check governed by the Parcel Update Frequency (In Configuration > Settings > Parcels)
By default, the time is one hour. So, perhaps by the time you read this, the the server will have detected your files and now the parcel is visible in the GUI.
06-10-2014 09:42 PM
Hi Thanks for prompt and detailed reply. It was very helpful. I will appreciate if you could further clarify on two points
1. I assume on agents we don't need to create local repo directory - it will get created as of host addition process by Cloudera Manager Admin as per config setting on Agent host.
2.Given that single Cloudera Manager can manage multiple clusters on the network and parcels will be ultimately stored on /opt/cloudera/parcels-repo on CM server What would be reason for creating a local web repository?
Are we looking at scenario where there could be mulitple CM servers (to serve redundancy )so that all of them could share single repo.