Elasticsearch is an open source search server project started by Shay Banon and published in February 2010. During this time, the project has grown into a major player in the field of search and data analysis solutions and is widely used in many more or lesser-known search applications. In addition, due to its distributed nature and real-time capabilities, many people use it as a document store.
Let's go through the basic concepts of Elasticsearch. You can skip this section if you are already familiar with the Elasticsearch architecture. However, if you are not familiar with this architecture, consider reading this section. We will refer to the key words used in the rest of the book.
Index is the logical place where Elasticsearch stores logical data, so that it can be divided into smaller pieces. If you come from the relational database world, you can think of an index like a table. However, the index structure is prepared for fast and efficient full-text searching, and in particular, does not store original values. If you know MongoDB, you can think of the Elasticsearch index as a collection in MongoDB. If you are familiar with CouchDB, you can think about an index as you would about the CouchDB database. Elasticsearch can hold many indices located on one machine or spread over many servers. Every index is built of one or more shards, and each shard can have many replicas.
The main entity stored in Elasticsearch is a document. Using the analogy to relational databases, a document is a row of data in a database table. When you compare an Elasticsearch document to a MongoDB document, you will see that both can have different structures, but the document in Elasticsearch needs to have the same type for all the common fields. This means that all the documents with a field called title need to have the same data type for it, for example, string.
Documents consist of fields, and each field may occur several times in a single document (such a field is called multivalued). Each field has a type (text, number, date, and so on). The field types can also be complex: a field can contain other subdocuments or arrays. The field type is important for Elasticsearch because it gives information about how various operations such as analysis or sorting should be performed. Fortunately, this can be determined automatically (however, we still suggest using mappings). Unlike the relational databases, documents don't need to have a fixed structure—every document may have a different set of fields, and in addition to this, fields don't have to be known during application development. Of course, one can force a document structure with the use of schema. From the client's point of view, a document is a JSON object (see more about the JSON format at http://en.wikipedia.org/wiki/JSON). Each document is stored in one index and has its own unique identifier (which can be generated automatically by Elasticsearch) and document type. A document needs to have a unique identifier in relation to the document type. This means that in a single index, two documents can have the same unique identifier if they are not of the same type.
In Elasticsearch, one index can store many objects with different purposes. For example, a blog application can store articles and comments. The document type lets us easily differentiate between the objects in a single index. Every document can have a different structure, but in real-world deployments, dividing documents into types significantly helps in data manipulation. Of course, one needs to keep the limitations in mind; that is, different document types can't set different types for the same property. For example, a field called title must have the same type across all document types in the same index.
In the section about the basics of full-text searching (the Full-text searching section), we wrote about the process of analysis—the preparation of input text for indexing and searching. Every field of the document must be properly analyzed depending on its type. For example, a different analysis chain is required for the numeric fields (numbers shouldn't be sorted alphabetically) and for the text fetched from web pages (for example, the first step would require you to omit the HTML tags as it is useless information—noise). Elasticsearch stores information about the fields in the mapping. Every document type has its own mapping, even if we don't explicitly define it.
Now, we already know that Elasticsearch stores data in one or more indices. Every index can contain documents of various types. We also know that each document has many fields and how Elasticsearch treats these fields is defined by mappings. But there is more. From the beginning, Elasticsearch was created as a distributed solution that can handle billions of documents and hundreds of search requests per second. This is due to several important concepts that we are going to describe in more detail now.
Elasticsearch can work as a standalone, single-search server. Nevertheless, to be able to process large sets of data and to achieve fault tolerance and high availability, Elasticsearch can be run on many cooperating servers. Collectively, these servers are called a cluster, and each server forming it is called a node.
When we have a large number of documents, we may come to a point where a single node may not be enough—for example, because of RAM limitations, hard disk capacity, insufficient processing power, and inability to respond to client requests fast enough. In such a case, data can be divided into smaller parts called shards (where each shard is a separate Apache Lucene index). Each shard can be placed on a different server, and thus, your data can be spread among the cluster nodes. When you query an index that is built from multiple shards, Elasticsearch sends the query to each relevant shard and merges the result in such a way that your application doesn't know about the shards. In addition to this, having multiple shards can speed up the indexing.
In order to increase query throughput or achieve high availability, shard replicas can be used. A replica is just an exact copy of the shard, and each shard can have zero or more replicas. In other words, Elasticsearch can have many identical shards and one of them is automatically chosen as a place where the operations that change the index are directed. This special shard is called a primary shard, and the others are called replica shards. When the primary shard is lost (for example, a server holding the shard data is unavailable), the cluster will promote the replica to be the new primary shard.
Elasticsearch handles many nodes. The cluster state is held by the gateway. By default, every node has this information stored locally, which is synchronized among nodes. We will discuss the gateway module in The gateway and recovery modules section of Chapter 7, Elasticsearch Cluster in Detail.
You may wonder how you can practically tie all the indices, shards, and replicas together in a single environment. Theoretically, it should be very difficult to fetch data from the cluster when you have to know where is your document, on which server, and in which shard. Even more difficult is searching when one query can return documents from different shards placed on different nodes in the whole cluster. In fact, this is a complicated problem; fortunately, we don't have to care about this—it is handled automatically by Elasticsearch itself. Let's look at the following diagram:
When you send a new document to the cluster, you specify a target index and send it to any of the nodes. The node knows how many shards the target index has and is able to determine which shard should be used to store your document. Elasticsearch can alter this behavior; we will talk about this in the Routing section of Chapter 2, Indexing Your Data. The important information that you have to remember for now is that Elasticsearch calculates the shard in which the document should be placed using the unique identifier of the document. After the indexing request is sent to a node, that node forwards the document to the target node, which hosts the relevant shard.
Now let's look at the following diagram on searching request execution:
When you try to fetch a document by its identifier, the node you send the query to uses the same routing algorithm to determine the shard and the node holding the document and again forwards the query, fetches the result, and sends the result to you. On the other hand, the querying process is a more complicated one. The node receiving the query forwards it to all the nodes holding the shards that belong to a given index and asks for minimum information about the documents that match the query (identifier and score, by default), unless routing is used, where the query will go directly to a single shard only. This is called the scatter phase. After receiving this information, the aggregator node (the node that receives the client request) sorts the results and sends a second request to get the documents that are needed to build the results list (all the other information apart from the document identifier and score).
This is called the gather phase. After this phase is executed, the results are returned to the client.
Now the question arises—what is the role of replicas in the process described previously? While indexing, replicas are only used as an additional place to store the data. When executing a query, by default, Elasticsearch will try to balance the load among the shard and its replicas so that they are evenly stressed. Also, remember that we can change this behavior; we will discuss this in the Understanding the querying process section of Chapter 3, Searching Your Data.