Creating a RAC cluster using Ansible (part 1)

In my previous blog post i explained how Vagrant can be used to create a 2-node VM environment for deploying a RAC cluster on it. In this post i want to dive into how to actually do the RAC install using ansible so you can easily create a RAC test environment on eg. your laptop.

I have created a git repo with the files that i am using for this blog post, this git repo can be found here

I am not going over over the basic of Ansible in this blogpost, i am assuming you have a basic understanding of Ansible and the YAML markup language. If you need a quick refresh on how Ansible works you can take a look here or here for example.

When trying to automate a RAC deployment you need to think about which steps to take and in what order. Also out-of-the-box Vagrant will do a serial deployment of your cluster which will for obvious reasons not work, the solutions for this is in my previous blogpost. It explains on how to setup Vagrant so ansible can do a “parallel” deployment. I will use a slightly modified version of the Vagrantfile i created in that blogpost.

If you go through the new vagrantfile you will see a few changes, the most obvious one are all the new parameters on top they can be used to easily tweak the RAC setup to your liking. You can change between things like CDB or not, memory settings, domain name etc. These parameters are all being passed over to ansible:

ansible.extra_vars = {
devmanager: "#{devmanager}",
db_create_cdb: "#{db_create_cdb}",
db_pdb_amount: "#{db_pdb_amount}",
db_name: "#{db_name}",
db_total_mem: "#{db_total_mem}",
gi_first_node: "#{host_prefix}1",
gi_last_node: "#{host_prefix}#{servers}",
domain_name: "#{domain_name}"
}

The final change i made to my vagrantfile is at the network section, i am providing the VM’s with a fix MAC address so i can refer to those later on when using udev as the device manager. I will get back to the MAC addresses later (in the next post) when we get to the part that configures device persistency

vb.customize ['modifyvm', :id, '--macaddress2', "080027AAAA#{rac_id}1"]
vb.customize ['modifyvm', :id, '--macaddress3', "080027AAAA#{rac_id}2"]

Vagrant is now able to create 2 VM’s that we can use for our RAC cluster. As you can imagine building 2 node RAC cluster with ansible will result to a rather large playbook that will be a bit cumbersome to work with. Also you might want to re-use certain parts later on for different deployments, in order to solve these issues Ansible has to ability to create roles where every role is defined set of actions all with its own set of files, templates and variables. For this RAC playbook i  have created a set of roles to handle specific parts of the installation:

  • common – setups some common linux setting we want to have no matter what
  • configure_network – setups linux network and configures stuff like DNS
  • configure_udev – If selected in the vagrantfile this will configure the system for udev
  • configure_asmlib – If selected in the vagrantfile this will configure asmlib
  • tools_install – installs a toolset with several scripts, install rlwrap etc.
  • rac_install_gi – installs and configures Oracle Grid Infrastructure
  • rac_install_db – installs and configures the Oracle Database home
  • rac_create_db – Creates the RAC database

The reason to use roles is to be keep a better overview of your ansible scripts but even more you can reuse roles for other playbooks. These roles are being called in the main Ansible file that is defined in the vagrantfile at line 88:

ansible.playbook   = "ansible/rac_gi_db.yml"

This YAML file kicks of the entire installation of our RAC cluster, if look in rac_gi_db.yml you will notice that there only a few lines there. This file is referencing the different roles. If Ansible is seeing role common, it expects that there is a subdirectory called roles, which should have a subdirectory in it with the role name, with another nested subdirectory tasks in it. In that subdirectory Ansible will look for a file called main.yml and runs that. So to visualize this, Ansible expects the following directory structure as a minimum for a role:

rac_gi_db.yml/roles/<em></em>/tasks/main.yml

The Ansible documentation itself has a more in-depth explanation on this. Another thing you might notice are the roles for configuring either asmlib or udev:

- { role: configure_udev, when: devmanager == "udev" }
- { role: configure_asmlib, when: devmanager == "asmlib" }

Like mentioned above, a role is a set of subdirectories that holds the information needed to perform the tasks of the specific role. Directories must have a specific name in order for Ansible to recognise them. Roles can have the following subdirectories, depending on your need.

  • defaults/ – where variables local this role are stored
  • files/ – needed if you want to upload files with ansible to a host. Eg. zipfiles for an Oracle GI install
  • tasks/ – This directory contains at least a file named main.yml which is holds all the tasks that this role has to execute.
  • templates/ – As the name suggests, it contains templates to be converted into files by Ansible. For Example a template for a responsefile that needs to be created with the correct database name and settings or a hosts file with the correct host names. These files are in the Jinja2 format.

If you look at line 11 in the vagrantfile you can see that there is a variable devmanager defined which is being parsed into Ansible at line 90. So depending on the value of the variable devmanager Ansible will either configure udev or asmlib. Global variables are furthermore set in the groupvars/all.yml file, these variables are being passed over too all roles in the group ‘all’.  In this ansible setup variables are being set at different levels, i made it so that the most important variables can be set in top of the vagrant file in lines 11 to 16 this can of course be expanded if needed. Besides passing variable direct to Ansible, specific roles can have their own variables, these are variables are set in rac_gi_db.yml/roles//defaults/main.yml. These are variables local to the execution of this specific role and can’t be referenced outside it. The final group of variables we need are host specific variable like ip address and network configurations in general. For these settings we have a file per host, were every host has its own file in the sub directory host_vars.

Now we know how Ansibles gets the parameters, we know what the roles are doing we can can actually start deploying. As soon as your start vagrant and it is done creating the initial VM’s (covered in my previous blogpost) it will start ansible and go through the roles and runs all the YAML files in the tasks directories. At one point you will find yourself in the need of getting some configuration files during installs but these files are dependent on eg. IP addresses, hostnames, database name etc. For example you can you use this task to create a dnsmasq config based on variables in an array called ether_result.results (which we created somewhat earlier in the task.

- name: Create dnsmasq configuration
template: src=dnsmasq.conf.j2 dest=/etc/dnsmasq.conf
with_items: "{{ ether_result.results }}"

In order to get these dynamic configurations we need to write these files with a language called Jinja2. You can take your normal configuration file and replace the variable with something that Jinja2 can work with. As an example let’s look at the dnsmasq.conf file in Jinja form:

# {{ ansible_managed }}
#/etc/dnsmasq.conf
domain-needed
bogus-priv

domain={{ domain_name }}
expand-hosts
local=/{{ domain_name }}/

server=10.0.2.3

interface=eth0
{% for msg in ether_result.results %}interface={{ msg.item.device }}
{% endfor %}

As you can see it is almost your regular dnsmasq.conf file but with some changes. At the top we have a variable  {{ ansible_managed }} which will be replaced with a string containing source file name, creation date and from which host it was created. The {{ domain_name }} variable is being passed on to ansible from vagrant and is actually set in the Vagrantfile.  the last lines of the files are a jinja2 loop going through whatever is in ether_result.results and finding every msg in msg.item.device, which will contain all the names of the network devices i have configured and placing it after the string “interface=”. After ansible is done with the task above you will end up with a file that looks something like this:

# Ansible managed: /my/path/t0/dnsmasq.conf.j2 modified on 2016-08-09 12:24:40 by klaasjan on exabook.local
#/etc/dnsmasq.conf
domain-needed
bogus-priv

domain=mydomain.local
expand-hosts
local=/mydomain.local/

server=10.0.2.3

interface=eth0
interface=eth1
interface=eth2

If everything is configured correctly, installation of the software can begin. So we need to move some zip files from the your host to the guest vm. Ansible does this with the copy command:

- name: Copy RDBMS zipfiles
copy: src={{ item }} dest={{ ora_stage }} owner={{ oracle_user }} group={{ oracle_group }} mode=0644
with_items: "{{ rdbms_zips }}"
when: inventory_hostname == "{{ gi_first_node }}" and rdbms_check.stdout != rdbms_home

The command above will copy the files in the variable {{ item }}, which is being read from the array {{ rdbms_zips }} on the line “with_items:”, it then copies these files to the whatever location is set in {{ ora_stage }} and with the correct user and group permissions. The final line means that it will only copy these files to the first node (we do not need them twice) and when there is not already and oracle home installed.

I mentioned above that ansible will look for the file item.rdbms_zips in the ./files subdirectory. So before installing make sure that all the needed files are there. To make this a bit easier for myself i have created a small script called rebuild.sh which will download these files for you (you need to have an MOS account). It will do so using Maris Elsin excellent getMOSPatch command, which is a jar files now and no longer a shell script. After downloading rebuild.sh will do what the name suggests and remove and rebuild your vagrant RAC environment. Use at your own risk, if it breaks a network config for you in VirtualBox or something in Vagrant… i have warned you.

In the next blogpost i will explain in somewhat more detail what the different steps are doing.

Advertisements

One thought on “Creating a RAC cluster using Ansible (part 1)

  1. Pingback: Creating a RAC cluster using Ansible (part 2) | Future Veterans

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s