Database Reference
In-Depth Information
ByteOrderedPartitioner, which orders the keys by their byte value, or the
CollatingOPP, which orders based on the U.S. English value. It is import-
ant to note that when using the ordered partitioners, the ability to do range
slices is given but may also lead to hot spots. If you change this paramet-
er, you will destroy all data in the data directories.
data_file_directories
Default: /var/lib/cassandra/data. Where Cassandra should store the data on
disk.
commitlog_directory
Default: /var/lib/cassandra/commitlog. Where Cassandra should store the
CommitLogs. This value should reside on a different volume from the
data_file_directories
.
disk_failure_policy
Default:
stop
. When a disk fails, Cassandra can react differently depend-
ing on need. By default, Cassandra will stop Gossip and Thrift on the
node, leaving it effectively dead. Optionally, you may specify
best_effort
, which will attempt to use the remaining good disks. The
best_effort
option is best when used with JBoD (“just a bunch of
disks”) configurations. You may also specify
ignore
, which will re-
spond with a failure when attempting to access the node.
saved_caches_directory
Default: /var/lib/cassandra/saved_caches. Specifies the directory in which
to store the saved caches.
commitlog_sync
Default:
periodic
. By default, Cassandra will acknowledge writes im-
mediately, then periodically fsync those writes to disk. Optionally, if
batch
is specified, Cassandra will not acknowledge writes until the
batch has been fsynced.
commitlog_sync_period_in_ms
Default: 10000. Specifies the time in which Cassandra will fsync writes to
disk. When using batch syncs, this value should be low as writes will
block until the sync happens.
commitlog_segment_size_in_mb
