Terraform Overview: It is the go-to tool for repeatable and easy infrastructure deployments. The author previously shared how to deploy Elasticsearch on AWS with Terraform and Packer and has since used it to deploy many clusters. BigData Boutique's offerings are cloud-agnostic and they help with projects on other clouds. All relevant code and documentation is on GitHub.

• Creating Immutable Images With Packer: To enable quick machine launches and avoid snowflake servers, images of servers are generated and used for provisioning. Packer is a tool for creating machine and container images. Two images are created: one for Elasticsearch node on latest Ubuntu and another with Kibana, Grafana, and Cerebro based on the first. More details in the README.
• Deploying an Elasticsearch Cluster With Terraform: Terraform is great for describing complex infrastructure easily and repeatably. After creating machine images with Packer, only some configurations need to be edited (machine sizes, nodes, location, SSH keys) and then terraform plan and terraform apply can be run to create the cluster. The recommended configuration is 3 master nodes, 2 data nodes, and 1 client node. Elastic's X-Pack is deployed with monitoring enabled but security disabled. Full details in the Terraform-Azure README.
• Client Nodes With Kibana, Grafana, and Cerebro: Once the cluster is deployed, client nodes expose Kibana instances, a Cerebro UI, and Grafana. Client nodes are password-protected and apps need to talk to them. Kibana initialization takes about 10 minutes.
• Elastic Discovery on Azure: The Azure "Classic" discovery plugin is deprecated. There is a PR for an Azure RM discovery plugin open for over a year with no progress. Defaulting to using vnet and naming conventions or file-based discovery is an option.
• Backups: The Azure repository plugin is installed on the cluster for index snapshots and restores. Official documentation is available.