Why is the Cassandra storage plugin not a highly available solution exactly?


Documentation for the Cassandra storage plugin states that it doesn’t support high availability, but no explanation on why exactly and I was wondering.

Is it because the plugin only supports hardcoded list of IP hosts and doesn’t run the network topology refresh to detect nodes based on host names instead of IPs?


High availability has a specific meaning here - that it is capable of supporting multiple Vault servers in a cluster, which elect an active node and manage which node is active via a lock in the storage.

Since Cassandra is all about eventual consistency, I can’t think of any way it could support the locking semantics required, and thus it can only support Vault in non-HA mode.