Production-ready maven project for Solr projects. The output of this build is an installation archive containing some handy defaults as well as some utility scripts.
You can setup solr configurations here:
Click here to see the Solr configuration folder
The build process generates a tarball suitable for installation. This archive contains a default config, some reasonable start/stop scripts, and a logging config. This is really intended only as a boilerplate - if your needs are different, you should either create a new assembly descriptor or just fork and customize this repo.
Uncompress the installation tarball somewhere, and use the start/stop scripts.
Start There are a number of params that can be sent to this startup script
- SOLR_PORT: the port for this instance of solr to listen.
- SOLR_DATA_DIR: path to the location for storing the index files.
Replication env params
- SOLR_MASTER: (true|false) if true, then this server is the master. If not true, then this server is a slave. It is best practice to always set this param.
- SOLR_MASTER_HOST: the host/post of the replication master
Examples:
Run solr on port 9000, store files in /var/solr/index0, set as master for replication
SOLR_PORT=9000 SOLR_MASTER=true SOLR_DATA_DIR=/var/solr/index0 sh start.sh
Run solr on port 9001, store files in /var/solr/index1, set as slave for replication, set master as the instance in the example above
SOLR_PORT=9001 SOLR_MASTER=false SOLR_MASTER_HOST=localhost:9000 SOLR_DATA_DIR=/var/solr/index1 sh start.sh
If you do not specific a port, your server will be running on the Solr default port 8983.
Solr Back Office - http://localhost:8983/solr
Stop
sh stop.sh
Solr replication is master-slave. http://wiki.apache.org/solr/SolrReplication
This project sets up hooks for the environment variables to accomplish a cluster. Leader election is manual right now. If the master goes down, then the data in slaves will get stale. To fix, a new master must be elected and services restarted on all nodes, then the data source should load into the new master. It is probably a good idea to setup a vhost for the index master and switch the pointer in the case of a failure so that upstream index data producers do not require an update to get the change.
Here is a diagram
