This is a Visual Studio project for creating an Elasticsearch cluster on Microsoft Azure using worker roles.

This is for people who want to run Elastic search on Azure in the Platform as a Service environment. This is also an opportunity to test Elasticsearch in a simulated distributed environment.

This is a visual studio project which can serve as a base for a solution based on an Elasticsearch cluster. The intent is to handle all the different aspects of setting up and managing a cluster

• Installation
• Configuration
• Plugin Setup
• Logging
• Snapshots
• Automatic Node Discovery
• Security

Typical usage involves installing the NuGet package on existing Web or Worker role in your visual studio project. Once the package is installed you update the configuration settings and include the service initializer in your WorkerRole.cs or WebRole.cs.

No, it runs in the full Azure Emulator on your desktop. The project is designed to work with azure files service for data and snapshots but falls back to a resource folder in the Azure Emulator. Other than that, there is no significant difference between running this project on the Azure Emulator and publishing it to Azure.

Install the NuGet package

Install-Package Elasticsearch-Azure-PAAS 

This package will add the required settings to any cloud service projects that refer to the Web or Worker Role the packaged was installed on. It also adds two folders called Config and Plugins. Please set the contents of these folders to always copy to output directory.

Config/elasticsearch.yml and Config/logging.yml are the base config files for elasticsearch. You can modify them if you want to add any settings of your own. The settings in these files will apply to all instances in the cluster.

There are a couple config steps before you can run it either in the Azure Emulator or in an Azure Cloud Service. The NuGet package has already added those settings with defaults from this project. Please change them where necessary.

The service will download the java jre installer from this url.

Default: http://127.0.0.1:10000/devstoreaccount1/installers/jre-8u40-windows-x64.exe

This tells the service whether JavaDownloadURL is a web accessible location or on the configured storage account

Default: storage Options: web, storage

This is simply the name used to save the jre installer into the download cache on the role instance

Default: jre-8u40-windows-x64.exe

The service will download the java jre installer from this url.

Default: https://download.elasticsearch.org/elasticsearch/elasticsearch/elasticsearch-5.3.0.zip

This tells the service whether ElasticsearchDownloadURL is a web accessible location or on the configured storage account

Default: storage Options: web, storage

This is simply the name used to save the elasticsearch package into the download cache on the role instance

Default: elasticsearch-5.3.0.zip

The service will use this connection string to download any download types marked as storage. It will also be used to create the share used to store elasticsearch data.

Default: UseDevelopmentStorage=true

This config value will be used to name the azure file service share. https://myaccount.file.core.windows.net/[ShareName]. This share will be used as a persistent store for elasticsearch data and snapshots.

Default: elasticdata

This is the drive letter assigned to the azure file service share on the role instance

Default: P:

This is the name of the endpoint elasticsearch nodes in the cluster will use to communicate which each other

Default: elasticsearch

If this option is enabled, the service will store data on the role instance rather than on the azure file service share. This might be handy when you need the maxium i/o speed and your data is easily replaceable. This is the only available option when using the emulator.

Default: true

This is the name of a container accessible through the storage account in the StorageConnection setting which contains plugin zip files you intend to install in your cluster.

Default: elasticsearchplugins

This is a pipe separated list of plugins. They will be installed using the built in plugin installer.

/bin/elasticsearch-plugin.bat install plugin-name

Sample: analysis-stempel|analysis-phonetic|analysis-smartcn

Once the package is installed and all configuration values are correct you can go ahead and initilize the service in WebRole.cs or WorkerRole.cs

    public class WorkerRole : RoleEntryPoint
    {
        private CloudStorageAccount storage = CloudStorageAccount.Parse(RoleEnvironment.GetConfigurationSettingValue("StorageConnection"));
        private ElasticsearchService service;
        public override void Run()
        {
            service.RunAndBlock();
        }

        public override bool OnStart()
        {
            // Not sure what this should be. Hopefully storage over smb doesn't open a million connections
            ServicePointManager.DefaultConnectionLimit = 12;


            var settings = ElasticsearchServiceSettings.FromStorage(storage);
            service = ElasticsearchService.FromSettings(settings);
            bool result = base.OnStart();

            Trace.TraceInformation("ElasticsearchRole has been started");

            return result;
        }

        public override void OnStop()
        {
            Trace.TraceInformation("ElasticsearchRole is stopping");

            service.OnStop();


            base.OnStop();

            Trace.TraceInformation("ElasticsearchRole has stopped");
        }


    }

If you find things a bit sluggish on startup in the emulator don't be alarmed. The code is written to use as much of the available resources as possible to minimize startup time. As a result, the initialization steps run concurrently using async tasks. After deployment to Azure, it will not re-run the initialization steps after the initial config. Therefore, subsequent role instance recycles will be much quicker.

The source of the NuGet package used to install this project on Web and Worker Roles is the Package.NuGet project located in this repository.

There are different options for configuring your cluster and other services on top of it. Here are a few ideas:

• Worker Roles only with public communication using Shield or private communication over a virtual network
• Worker Roles for elasticsearch and separate Public facing Web Roles which use elasticsearch as a backend service
• Public facing WebRoles which run both iis and Elasticsearch

I hope this project is useful to you. If you have a quick question and don't want to create an issue you can reach me on twitter @garvincasimir.