Ambari server installation fails at Confirm Hosts

[#1]

Hi all,

My installation of Amabari server fails at the Confirm Hosts step at the registration phase. I have 4 hosts I want to use for this cluster. Of the 4 hosts, hadoop02 to hadoop04 successfully complete the registration step. The failure comes at hadoop01 when it tries to complete registration. So the whole step of Confirm Hosts fails.

Hadoop01 is the one I intend to use as the master, and that's where I'm running the GUI from on 8080. I'm using the 'root' user to ssh to all the hosts. I also tried using the 'hadoop' user to to ssh to all the hosts. I've verified that I can do that to all hosts (including hadoop01) by ssh'ing into each host as those users by hand.

This is what I have in the logs when hadoop01 decides to fail:

INFO:root:BootStrapping hosts ['hadoop01.example.com',
 'hadoop02.example.com',
 'hadoop03.example.com',
 'hadoop04.example.com'] using /usr/lib/python2.6/site-packages/ambari_server cluster primary OS: redhat7 with user 'root' sshKey File /var/run/ambari-server/bootstrap/4/sshKey password File null using tmp dir /var/run/ambari-server/bootstrap/4 ambari: hadoop01.example.com; server_port: 8080; ambari version: 2.1.0; user_run_as: root
INFO:root:Executing parallel bootstrap
WARNING:root:Bootstrap at host hadoop01.example.com timed out and will be interrupted
INFO:root:Finished parallel bootstrap
30 May 2016 18:43:12,461  INFO [pool-9-thread-1] BSHostStatusCollector:55 - Request directory /var/run/ambari-server/bootstrap/4
30 May 2016 18:43:12,461  INFO [pool-9-thread-1] BSHostStatusCollector:62 - HostList for polling on [hadoop01.example.com, hadoop02.example.com, hadoop03.example.com, hadoop04.example.com]
30 May 2016 18:43:12,909  INFO [pool-9-thread-1] BSHostStatusCollector:55 - Request directory /var/run/ambari-server/bootstrap/4
30 May 2016 18:43:12,910  INFO [pool-9-thread-1] BSHostStatusCollector:62 - HostList for polling on [hadoop01.example.com, hadoop02.example.com, hadoop03.example.com, hadoop04.example.com]

Can I get some help in getting past this step?

[#2]

Check if you can passwordless ssh to the same host (hadoop1) using the key. I believe you set this from hadoop1 to connect to hadoop2, hadoop3 and hadoop4 but not hadoop1 itself.

If this does not work, you can try manual registration of ambari using http://docs.hortonworks.com/HDPDocuments/Ambari-2.2.2.0/bk_ambari_reference_guide/content/ch_amb_ref...

[#3 · Accepted solution]

@Tim Dunphy

1. Make sure your below command has same output -

$hostname

$hostname -f

2. Make sure if iptables and selinux are disabled

3. SSH to localhost and make sure it works without password.

4. Once you make sure your passwordless ssh is working you try installing agent manually also - http://docs.hortonworks.com/HDPDocuments/Ambari-2.2.1.0/bk_ambari_reference_guide/content/ch_amb_ref...

[#4]

Hey guys, Thanks for your help! I was able to resolve it. The problems were that trying to complete the installation as the hadoop user didn't have enough permissions to perform all the actions it needed. That was problem one! Using the root user to ssh into all the hosts worked. But only after I had opened up ports 8440 and 8441 using security groups did the installation phase complete successfully. Sorry I neglected to mention that this was being setup in AWS in the OP. My bad! Also IP tables was off as the ports are being managed by AWS security groups. Unfortunately, the Ambari docs are a bit sparse. And, unless I've missed something, don't include info on what ports to open. If I'm right about that, this sucks and needs to be corrected. Thanks

[#5]

After changing the hostname in the ambari configuration file restart the ambari server.