See all
Edit on GitHub

Out of the box Virto Commerce uses Elastic Search. Elastic search is installed as a windows service when running on-premises and as a worker role when running in azure environment. It run is a tomcat java container.

Search Indexing

Indexing can be done for different types of documents, called DocumentType's in the code. Each document type indexing in turn is split into 3 parts:

  1. Create partitions - splits the task of indexing large data set into sets, by default of 100 items.
  2. Create documents - creates name/value documents that can be then submitted to the indexing server.
  3. Publish document - publishes the documents to the indexing server.

The indexing process can be done in parallel on multiple servers aka WorkRole or WebRole in Azure Environment.

Create Partitions

During this process we create partitions for the set of data to be indexed. It can be based from the full set of data like catalog products or a subset based on the last date the indexing ran. The partitions are then added to the Queue. The partition creator class must implement ISearchIndexPartitionCreator interface.

Create Documents

During this process we dequeue messages from the queue where partitions were submitted to and create name/value documents. Class that implement this functionality must implementISearchIndexDocumentCreator interface where T is typically "string[]". The class will be passed an array of strings which contains ID's of objects that will need to be converted into documents.

Publish Documents

As soon as the first set of documents is created the publishing starts. The documents are submitted using the provider interface and are handed over to the default search provider which currently can be either Lucene or SOLR. The document publisher implements ISearchIndexDocumentPublisher interface.

Indexing Coordination

The overall process is controlled by the class implementing ISearchIndexController. For example AzureSearchIndexController does that for the azure environment. That class puts messages into the queue, dequeues items and saves the progress to the BuildSettings table. It also makes sure the task of creating partitions is only executed once across all servers.


Each document type needs to have an IndexBuilder configured. For example Catalog has a CatalogItemIndexBuilder configured in search.config file. It needs to implement ISearchIndexBuilderinterface which simply contains methods to initiate 3 parts of the indexing process: CreatePartitions, CreateDocuments, PublishDocuments and is instantiated by the ISearchIndexController.