Cassandra (Helm)
What does zop.dev create for Cassandra (Helm)?
zop.dev deploys Apache Cassandra inside the cluster with the Bitnami Helm chart, attached through a namespace's cassandra_db map, where replica_count defaults to 1 and persistence_size to 10 GB. The bill is your own nodes and disks, and 1 replica means no quorum and no tolerance for losing a single pod.
| Field | Value |
|---|---|
| Cloud | cross-provider |
Apache Cassandra is a distributed wide-column database for high-write workloads. zop.dev deploys it in-cluster via the Bitnami Helm chart on any supported cloud, attached to a namespace.
Bitnami Cassandra attached via cassandra_db
Bitnami Cassandra Helm release inside the cluster, attached through the namespace’s cassandra_db map.
Replica count and persistence size
name; admin_user; replica_count default 1; persistence_size default 10 GB.Terraform only, and not on OCI namespaces
Cassandra exists in the Terraform layer only. It is not a datastore kind in the backend-api catalog, and OCI namespaces do not support it.
One replica is a development posture
Cassandra’s whole design assumes a ring of peers. Its consistency guarantees come from
quorum, and quorum needs at least 3 replicas. The default replica_count of 1 collapses that
to a single point of failure: one pod eviction, one node drain, and the database is down with
it. That default is right for development environments where the data is disposable, and wrong
for anything whose data is not. Raising the count is a capacity decision too: each replica is
a JVM with real memory appetite, multiplied across your nodes.
Disks outlive pods
Each replica claims a persistent volume at persistence_size (10 GB by default), and those
PVCs have their own lifecycle. A restarted or rescheduled pod reattaches its volume, which is
the durability story, but volumes bill as cloud disks for as long as they exist, and deleting
the Helm release does not necessarily delete the claims. Growth only goes one way: volumes can
usually be expanded but never shrunk, so the 10 GB default is a floor that ratchets upward.
Teardown hygiene means checking for orphaned PVCs after the release is gone.
A Terraform-layer resident
Cassandra exists only in the Terraform layer. It is not a datastore kind in the backend-api
catalog, and OCI namespaces do not support it. Practically, that means it is attached through
a namespace’s cassandra_db map (with name, admin_user, replica and persistence settings)
rather than appearing in the application catalog the way managed databases do. Teams get an
in-cluster wide-column store for high-write workloads without a managed-service bill. The
corresponding trade is that operations, upgrades, and capacity are theirs to own.