Community Articles

Find and share helpful community-sourced technical articles.
Announcements
Celebrating as our community reaches 100,000 members! Thank you!
avatar
Super Guru

Since getting started with Hadoop & ELK I have taken the original Hortonworks 5.x Elasticsearch and Kibana Management Pack for HDP 2.6.5 and upgraded it to ELK 6.3.2.  During this project I also added Logstash, FileBeat, and MetricBeat.  Next, I upgraded the Management Pack to work with HDP 3.x and HDF 3.x.  

 

In this article I am going to share with you all of the steps required to upgrade the ELK 6.3.2 Management Pack to latest, greatest versions of ELK 7.4.2.  You can find all of the files in my GitHub repo: DFHZ ELK Mpack

 

First, I create a known working test cluster making sure the 3.4-0 management pack is operational and my test environment is suitable.

Commands for a single node HDF test cluster:

# history
    1  hostnamectl set-hostname elk.cloudera.com
    2  yum install nano wget -y
    3  nano /etc/sysconfig/selinux
    4  nano /etc/cloud/cloud.cfg
    5  nano /etc/hosts
    6  reboot
    9  wget -O /etc/yum.repos.d/ambari.repo http://public-repo-1.hortonworks.com/ambari/centos7/2.x/updates/2.7.3.0/ambari.repo && yum --enablerepo=extras install epel-release -y && yum install nano java java-devel ambari-server ambari-agent -y && ambari-server setup -s && ambari-server install-mpack --mpack=http://public-repo-1.hortonworks.com/HDF/amazonlinux2/3.x/updates/3.4.1.1/tars/hdf_ambari_mp/hdf-ambari-mpack-3.4.1.1-4.tar.gz && ambari-server start && ambari-agent start
   10  ssh-keygen
   11  cat ~/.ssh/id_rsa
   12  cat ~/.ssh/id_rsa.pub
   13  nano ~/.ssh/authorized_keys
   14  ssh root@elk.cloudera.com
   15  ambari-server install-mpack --mpack=https://github.com/steven-dfheinz/dfhz_elk_mpack/raw/master/elasticsearch_mpack-3.4.0.0-0.tar.gz --verbose
   29  ambari-server restart
   35  python /var/lib/ambari-server/resources/scripts/configs.py -u admin -p admin -n ELK -l elk.cloudera.com -t 8080 -a set -c cluster-env -k  ignore_groupsusers_create -v true

 ** note: before #35, complete the Cluster Install Wizard with base components (Zookeeper & Ambari Metrics).  When Base Cluster is done, run the python command and install ELK Via Ambari Add Service Wizard.  Without python command, ELK stack install will fail on user/group issues.

 

Second, I download then unpack the management pack above.  I make a new version and begin to edit the file structure for "7.4.2".  My edits were as follows:

  1. Rename and version the archive folder elasticsearch_mpack-3.4.0.1-0
  2. Change all component folders from 6.3.2 to 7.4.2
  3. Update mpack.json from 6.3.2 to 7.4.2
  4. Update Component File Set (Find/Replace) all 6.3.2 to 7.4.2
  5. Update addon-services/ELASTICSEARCH/7.4.2/repos/repoinfo.xml to 7.x
  6. Create a new archive elasticsearch_mpack-3.4.0.1-0.tar.gz

In this stage I just want to make sure that the stack versions are coming over correctly.  I update the GitHub repo, uninstall the management pack, restart ambari, install the new management pack, restart ambari, and go to the Add Service from ambari.   I can now see my ELK components reporting version 7.4.2.    I complete the install to make sure everything installs, but I do not expect the services to run.  There are still quite a few things to do to each components .yml config files.

Third, I start to work with Elasticsearch and Kibana config files.   For these steps I need multi node cluster in order to have an Elasticsearch Master and a Slave.   I use two local vagrant nodes to complete this task.  First I spin up a single node, add the ELK 7.x repos, manually install each component, and grab a copy of the 7.x .yml files (elasticsearch.yml, logstash.yml, kibana.yml, filebeat.yml, metricbeat.yml).   I save these for later.

Vagrant Commands For Manual Elk Install:

[root@elasticsearch]# history
1 cd /etc/yum.repos.d
2 yum install nano -y
3 nano elastic.repo
4 yum install elasticsearch logstash kibana filebeat metricbeat -y
5 cat /etc/elasticsearch/elasticsearch.yml
6 cat /etc/logstash/logstash.yml
7 cat /etc/kibana/kibana.yml
8 cat /etc/filebeat/filebeat.yml
9 cat /etc/metricbeat/metricbeat.yml

 Next I spin up two nodes, an Ambari-Master with Agent, and another Agent only node.

Master Server Commands:

[root@c7302 ~]# history
    1  wget -O /etc/yum.repos.d/ambari.repo http://public-repo-1.hortonworks.com/ambari/centos7/2.x/updates/2.7.3.0/ambari.repo && yum --enablerepo=extras install epel-release -y && yum install nano java java-devel ambari-server ambari-agent -y && ambari-server setup -s && ambari-server install-mpack --mpack=http://public-repo-1.hortonworks.com/HDF/amazonlinux2/3.x/updates/3.4.1.1/tars/hdf_ambari_mp/hdf-ambari-mpack-3.4.1.1-4.tar.gz && ambari-server start && ambari-agent start
    2  ambari-server install-mpack --mpack=https://github.com/steven-dfheinz/dfhz_elk_mpack/raw/master/elasticsearch_mpack-3.4.0.1-0.tar.gz --verbose
    3. python /var/lib/ambari-server/resources/scripts/configs.py -u admin -p admin -n ELK -l c7302.ambari.apache.org -t 8080 -a set -c cluster-env -k  ignore_groupsusers_create -v true

Agent Server Commands:

[root@c7303 vagrant]# history
    1  wget -O /etc/yum.repos.d/ambari.repo http://public-repo-1.hortonworks.com/ambari/centos7/2.x/updates/2.7.3.0/ambari.repo && yum --enablerepo=extras install epel-release -y && yum install nano java java-devel ambari-agent -y && mbari-agent start

Fourth,  I complete Cluster Install Wizard for my base cluster. I then begin comparing the original Fileset config files with the new .yml I have saved. Inside of each component folder I rename the old .xml to .xml.6.3.2. I then make a new .xml with the new 7.x file contents. Next, I convert the required values, parameters, and settings using the variables set up in params.py. Some variables changed names completely. Most remained as is. Overall there are not many differences so the translation from 6.x to 7.x was pretty easy. If you need more detail here unpack the different versions in the GitHub repo and compare the filesets. Feel reach to reach out directly via a message or send a comment if you have questions here too.  

To complete the config translation from 6.x to 7.x I worked with each component one at a time in a methodical process:

  1. Complete Required .yml Changes in Fileset
  2. Commit to GitHub
  3. Uninstall/Re-Install the management pack using full GitHub Url
  4. Restart Ambari
  5. Add Service & Test
  6. Adjust .yml on Node (until service runs as expected)
  7. Stop & Remove Service From Ambari
  8. Repeat

 Install:

ambari-server install-mpack --mpack=https://github.com/steven-dfheinz/dfhz_elk_mpack/raw/master/elasticsearch_mpack-3.4.0.1-0.tar.gz --verbose
ambari-server restart

Uninstall:

ambari-server uninstall-mpack --mpack-name=elasticsearch-ambari.mpack
ambari-server restart

Last, I reset my entire cluster, and complete a full Base Cluster Install. Next, install the management pack, install the full ELK Stack and confirm there are no issues.   To be thorough I also needed to complete this entire test in the HDP cluster too.   During the work, I also identified some questionable configs and could easily spend many more days on this M-Pack.  However, these are the only things required to make it work and allow deeper configuration changes within Ambari.

 

When complete you should have a fully working cluster with ELK:

Screen Shot 2019-11-21 at 9.43.19 AM.png

Notes and Lessons Learned:

  1. Work in small test-able loops
  2. Always restart Ambari-Server after making management pack changes
  3. Do not allow your editor or operating system to add/files inside of the management pack fileset.  Example: DS_Store
  4. Do not make accidental character changes in fileset which could invalidate XML or Python
  5. Be careful of characters in .yml file content which invalidate XML. Example: <user>:<pass>
  6. If things are not behaving as expected, blow away your current test-loop, and start a fresh one using a previous step.
  7. Always keep old version of important files. 
997 Views
0 Kudos