Operations

Backups

We have replication, but this does not protect against bad data. Only against data loss in case of failure.

We still need backups in case a rollback is necessary.

Use nodetool.

Restores

Restoring a backup requires shutting down the node to be restored before restoring.

Use sstableloader tool.

Incremental Snapshots

Incremental snapshots of data can consume large amounts of disk space, so remember to delete old ones.

This can be configured in the Cassandra yml config file.

Compression

Cassandra uses compression for performance (at the cost of CPU usage). There are various types of compression that can be used:

  • Google Snappy (faster reads/writes);
  • DeflateCompressor (Java zip, slower but better compression); and
  • LZ4Compression (only in 1.2.2 or later).

Compression is applied on a per-table basis.