Added by Avinash Lakshman, last edited by Avinash Lakshman on Mar 21, 2009
()
There is a standard config that comes with the download of Cassandra. Here is a description of some of the basic information that needs to be in the configuration file:
| Property | Description | Default |
|---|---|---|
| ClusterName | The name of this cluster. This is mainly used to prevent machines in another logical cluster from joining any other cluster. | Test |
| HashingStrategy | Partitioning strategy required. Choices are RANDOM and OPHF | RANDOM |
| RackAware | Replication policy preferred. For now just say false | false |
| MulticastChannel | This is not used anymore. | |
| ReplicationFactor | Replication factor desired | 3 |
| ZookeeperAddress | This is not used currently. | |
| RpcTimeoutInMillis | RPC timeout in milliseconds | 2000 |
| StoragePort | TCP port on which all inter cluster communication takes place. | 7000 |
| ControlPort | UDP port on which all cluster membership communications occur. | 7001 |
| ColumnIndexSizeInKB | Size after which columns are indexed by the system | 256KB |
| HttpPort | HTTP port for the WEB dashboard. | 80 |
| MetadataDirectory | Cassandra metadata information location. | |
| CommitLogDirectory | Location of the commit log. | |
| CommitLogRotationThresholdInMB | Size after which commit log needs to be rotated. | 128 |
| Seeds | This is for dynamic node discovery. Nodes discover other nodes by initially contacting the seeds. |
After the installation step navigate to the directory that contains the ANT build config and type "ant jar" at the prompt. Some common problems in the deployment steps are:
- Modify the start-server script to the -Dstorage-config property to the location which contains the storage-conf.xml
- Make sure the log4j.properties also resides in the same location as above.
- Make sure the directory location for the log4j log files are pre-created or you could make them part of the start-server script.
- Modify the Java heap settings in the start-server as per your requirements.
Typical Cassandra deployment has been on machines with 16GB RAM dual CPU quad core machines and more disks the better. We have typically used 3 disks per machine.