Deploying AEM Solr Search
TODO
AEM Solr Search provides a Solr quickstart distribution using Maven. It is available in the aemsolrsearch-quickstart sub module. This sub project was designed with the following goals:
- Provide an easy-to-use Solr runtime environment that can be built purely from a single Maven POM. Currently, it fetches the Solr 4.x WAR file and deploys it to a Jetty container.
- Provide a sample Solr collection that works with the AEM Solr Search Geometrixx custom extension.
- Provide an example how to manage a Solr project using source control.
- Provide a rapid development environment. The development cycle is as simple as editing the Solr home files and running the Maven build. Subsequent changes are as simple as editing the Solr home files, stopping the Jetty container (
CTRL+C) and re-running the Maven build.
While this sub project is still supported for development use, it will only be maintained for Solr 4.x If you need Solr 5 support, refer to the Docker section below.
It should also be noted that the quickstart distribution should only be used for local development or in conjunction with your continuous integration process. This should not be used in production or in lower environments (i.e., dev, QA, pre-production, etc.).
A number of Docker images are available for the Solr runtime. Like the quickstart distribution, they provide sample Solr runtime environments using the AEM Solr Search custom extension Geometrixx sites. At the time of this writing four versions are available:
- Solr 4 - Standalone
- Solr 4 - SolrCloud
- Solr 5 - Standalone
- Solr 5 - SolrCloud
Pre-built images are available on DockerHub.
For those interested in how the Docker images are built, the Dockerfiles are available in the aemsolrsearch-docker sub project on GitHub. These Dockerfiles serve as a good reference for DevOps engineers who need step-by-step instructions on how to deploy Solr in non-production environments.
TODO
The deployment diagram below is recommended for most production deployments; however, the number of ZooKeeper and SolrCloud nodes may vary depending on your needs. In addition to running SolrCloud, this architecture assumes two separate collections for AEM; one for author and for publish.
For your Solr deployment refer to Taking Solr to Production.
The deployment diagram below is recommended if you are planning on running Solr in Standalone mode. In this model, both author and publish will have separate cores with indexing only occurring on the master. The slaves should handle all queries.
