classicvalues / terraform-ceph-installer

Terraform Installer for CEPH on Oracle Cloud Infrastructure

Terraform Installer for Ceph in Oracle Cloud Infrastructure

Version: 1.1

About

The scripts in this repository allows you to streamline and/or replicate your Ceph deployment in Oracle Cloud Infrastructure (OCI).

Ceph is an open source distributed storage system designed for performance, reliability and scalability. It provides interfaces for object, block, and file-level storage. Ceph is now widely used and fully supported on Oracle Linux as described in the Release 3 of Ceph Storage for Oracle Linux.

Terraform is an Open Source Software (OSS) for building, changing, and versioning Cloud infrastructure safely and efficiently. Terraform Provider for OCI allows one to create the necessary Infrastructure resources and configure them in OCI.

The Terraform Installer for Ceph provides Terraform scripts for installing Ceph Clusters in Oracle Cloud Infrastructure (OCI). It consists of a set of Terraform scripts and modules, bash scripts, and example configurations that can be used to provision and configure the resources needed to run a Ceph Storage Cluster on OCI.

Ceph Cluster Configuration Overview

A typical Ceph Cluster includes multiple virtual or bare metal machines, referred to as nodes, serving the role of one (or more) of the following:

• Deployer - for installing Ceph on all other nodes
• Monitor - for maintaining the maps of the cluster state and authentication - typically 2 or 3 for high availability
• Manager - for keeping track of the cluster state and exposing the cluster information
• Object Storage Daemon (OSDs) - for storing and handling data - typically many of them for redundency
• Metadata Server (MDS) - for storing metadata on behalf of the Ceph Filesystem

Deploying Ceph involves creating infrastructure resources (e.g., compute, network, storage), setting them up for Ceph installation, installing various packages on all machines, and finally configuring and deploying the cluster. This requires a fair bit of knowledge about OCI and Ceph. Carrying out the entire process manually is tedious and error prone.

However, by using the sctipts in this repository, you can create the necessary infrastructure resources and deploy a Ceph cluster using those resources. The behavior of the scripts are controlled by various configuration files. By changing the variables defined in these files, you can control what resources are created in OCI (e.g., the number and type of VMs to create for various Ceph nodes), how the Ceph Cluster is configured (e.g., the level of replication), etc.

Creating a Ceph Cluster

This README guides you through the following steps to accomplish the goal of creating a cluster shown in the pcture below:

• Install Terraform
• Download the scripts
• Setup for access to OCI
• Customize the Ceph Cluster
• Execute the scripts
• Login to a Ceph Admin node and check the status of the Cluster

Prerequisites

Designate a machine to run Terraform. This machine needs to have the credentials (e.g., the .pem key files under ~/.oci) to access the your tenant in OCI.

This machine should also have the rsa key pair generated for the Linux user on this machine. This key will be supplied during the creation of the compute nodes to allow the Linux user to perform password-less ssh logins the newly created VMs.

If not, generate the key pair using:

$ ssh-keygen -t rsa

Install Terraform

Add and/or enable the following yum repository:

[ol7_developer]
name=Oracle Linux $releasever Development Packages ($basearch)
baseurl=https://yum.oracle.com/repo/OracleLinux/OL7/developer/$basearch/
gpgkey=file:///etc/pki/rpm-gpg/RPM-GPG-KEY-oracle
gpgcheck=1
enabled=1

Then execute:

$ yum install terraform
$ yum install terraform-provider_oci

Download Scripts

Download the Terraform Ceph scripts from this repository.

$ git clone https://github.com/oracle/terraform-ceph-installer.git
$ cd terraform-ceph-installer

Setup for access to your OCI tenant

Copy and edit the sample environment file to fill in the information particular to your tenancy which includes the tenant, region, user id, and the credentials to access OCI. This is the same information you would require to use the OCI-CLI. Please refer to Required Keys and OCIDs to find out the OCIDs and generate the keys needed.

Source it to export the variables in the file.

$ cp env-vars.sample env-vars
$ vi env-vars
$ . env-vars

Customize for your OCI tenant

Create a copy (with a .tf extension) of one of the given examples and modify to fit your need and your environment. The example in variables.ex1 assumes that have existing VCN and subnets for your tenant in OCI and you deploy the cluster using those subnets.

$ cp variables.ex1 variables.tf

You will definitely need to edit the following variables:

• existing_vcn_id - The ocid for the existing VCN. You can find it by login into you tenant using a browser.
• existing_subnet_ids - The list of ocids for the existing subnets. You can find them by login into you tenant using a browser. You may also need to edit the following variables:
• instance_os - The full name of the operating system that you want on all nodes. You can find it by login into you tenant using a browser. Use the latest availble one. Follow the same format included in the example.
• instance_shapes - The shapes for the compute resources. Default is VM.Standard1.2 for all nodes. You can use different shapes for differnt types of nodes, i.e., deployer, monitor, osd, mds, and client.
• create_volume - The default is true which means it will create a block storage for each OSD. If you choose a shape with NVMe drives for OSDs, change it to false.

Execute Scripts

# Initialize your Terraform configuration including the modules
$ terraform init
# See what Terraform will do before actually doing it
$ terraform plan
# Provision resources and configure the Ceph cluster on OCI
$ terraform apply
### Check Cluster Status

Upon the successful completion, the scripts will print out the names and IP addresses for all the compute nodes which can then be used to access the nodes. The tail end of the output will look like the following:

Outputs:
ceph_client_hostname = [
    test-ceph-client
]
ceph_client_ip = [
    100.100.45.31
]
ceph_deployer_hostname = test-ceph-deployer
ceph_deployer_ip = 100.100.45.30
ceph_mds_hostname_list = [
    test-ceph-mds-0
]
ceph_mds_ip_list = [
    100.100.45.49
]
ceph_monitor_hostname_list = [
    test-ceph-monitor-0,
    test-ceph-monitor-1
]
ceph_monitor_ip_list = [
    100.100.45.33,
    100.100.46.30
]
ceph_osd_hostname_list = [
    test-ceph-osd-0,
    test-ceph-osd-1,
    test-ceph-osd-2,
    test-ceph-osd-3
]
ceph_osd_ip_list = [
    100.100.45.33,
    100.100.45.32,
    100.100.46.31,
    100.100.48.30
]

If you need to list the IP addresses again in the future, enter:

$ terraform show

Check the status of the Cluster

$ ssh -l opc <client-ip-address>
$ ceph status
$ df -h

Known issues and limitations

• The scripts doesn't check for the validity of the input variables or any inconsistencies among them. Terraform will execute but fail if the compute shape or image doesn't exist in your environment or the specified subnet id for a compute node doesn't belong to the specified availability domain for the same node. It is your responsibility to make sure the inputs are valid and consistent with one another.
• Uppercase letters on network resource names may cause problems.