A new loadbalanced, failover capable SolrJ client

The solr-extensions-client is a new implementation of the LBHttpSolrServer shipping with Solr. It is designed to be open for extension and enables a transparent request failover.

The LBHttpSolrServer does a great job when it comes to balance requests across a bunch of Solr servers. But unfortunately it is not designed to be open for extension, which this implementation tries to solve.

The FailoverLoadbalancedHttpSolrServer extends the SolrServer class and exposes the following capabilities

Load balancing strategy
Exception handling strategy
Instance healthchecks

Installation

Add the following dependency to your projects pom.xml (users of other build tools should know how to translate the following):

<dependency>
   <groupId>org.apache.solr.extensions</groupId>
   <artifactId>solr-extensions-client</artifactId>
   <version>0.1.3</version>
</dependency>

Because until now the jar is not yet available in maven central you need to add the following repository to your pom.xml as well:

<repositories>
   <repository>
      <id>bintray</id>
      <name>bintray</name>
      <url>http://dl.bintray.com/magro/maven</url>
   </repository>
</repositories>

Usage

// create a new client
FailoverLoadbalancedHttpSolrServer server = new FailoverLoadbalancedHttpSolrServer(
   httpClient,
   new BinaryResponseParser(),
   new RoundRobinSolrLoadBalancerStrategy(),
   new ByErrorRateSolrExceptionHandler(),
   new PingAwareSolrServerAliveChecker(5, 10, TimeUnit.SECONDS));

// start adding solr urls
server.addSolrServer("http://some.solr.url/solr");

Configuration

The project contains some utility classes:

ExplanatoryHttpSolrServer – does a proper to string for the HttpSolrServer
HttpClientSolrConfigurationBean – configures a http client with some Solr defaults. Use in IoC containers like Spring.

`SolrLoadBalancingStrategy`

Default load balancing strategy is the RoundRobinSolrLoadBalancerStrategy that distributes requests round robin across the configured servers. Dead servers are skipped and a exception is thrown if no alive Solr server is left.

When issueing more than one request to a Solr server, it might be handy to send those to the same Solr instance. If you for example request available facets for a given query in a first request and retrieve matching documents in a subsequent request, query time will decrease when working on the same facet/filter/document caches.

To achieve this, the StickyDelegatingSolrLoadBalancerStrategy utilizes the Spring RequestContextHolder to store the first Solr instance used in the current request context. The following requests are sent to the same Solr instance.

`SolrServerAliveChecker`

PingAwareSolrServerAliveChecker – iterates through all configured servers and checks their availability via a ping request

DeadSolrServerAliveChecker – iterates through all servers marked as dead and checks their re-availability via a ping request

`SolrExceptionHandler`

ByErrorRateSolrExceptionHandler – disables a Solr instance if the thrown exception rate exceeds the configured error rate per minute

Extend

Contributions are welcome! We'd like to see async http client support and more sophisticated lb strategies.

Build

You need to haven Maven installed. Check out the project, run

$ mvn clean verify

You'll find the client jar in the target directory.

Release

In order to do a release we have to prepare the release

$ mvn release:prepare

The parameter developmentVersion can be used to set the new version of your local working copy. Afterwards we perform the release (to deploy to bintray add -Pbintray)

$ mvn release:perform

License

Apache License

netboynb/solr-extensions-client