I'm running into issues setting up cloudera manager for cdh on a cluster. I've set up the head node to run cloudera manager and am trying to add compute node hosts to it. It wants to add the hosts, but fails the install. I followed the instructions to set up an http server to host the install repository, as my compute nodes don't have access to the internet. I was looking through the logs and found that it was trying to download files from internal http servers using my head node's FDQN, but the files it was looking for were actually at an address with simply the host name. For example, at one point it was trying to download a file from http://myhostname.us.my.domain:8091. Going to this adress on the compute node host does not connect. However, going to http://myhostname:8091 does connect, and goes to the file that it is looking for. How do I make the install files start using the correct addresses? This may come down to an error with my internal DNS server but I am not sure. Anyone know if this is an issue with my local DNS server or if there is something I can do in cloudera manager?
I think it is best that you correct your name resolution to account for domain as this sort of thing can cause problems for hadoop (especially with Kerberos/TLS involved).
You don't necessarily have to use /etc/hosts if you are using a DNS server for resolution, but just make sure that everything is configured to use FQDN.
For now, if you are having trouble just accessing the http server, perhaps try using IP address rather than host name in the Remote Parcel Repository URLs config.