Hex.pm

Note: This library is under active development and the API may change.

AshScylla

An Ash Framework data layer for ScyllaDB/Apache Cassandra

Quick StartFeaturesDocumentationContributingLicense


Overview

AshScylla enables you to use ScyllaDB or Apache Cassandra as a persistence layer for your Ash Framework resources. It implements the Ash.DataLayer behaviour using Xandra (a native Elixir CQL driver) to communicate via CQL (Cassandra Query Language).

Key Benefits


Quick Start

Prerequisites

Installation

Add ash_scylla to your dependencies in mix.exs:

def deps do
[
{:ash_scylla, "~> 0.6.0"}
]
end

Minimal Setup

1. Configure a Repo:

# lib/my_app/repo.ex
defmodule MyApp.Repo do
use AshScylla.Repo,
otp_app: :my_app
end

2. Configure the Repo in config/config.exs:

config :my_app, MyApp.Repo,
nodes: ["127.0.0.1:9042"],
keyspace: "my_app_dev",
pool_size: 10

3. Add the Repo to your supervision tree:

# lib/my_app/application.ex
children = [
MyApp.Repo,
# ...
]

4. Generate a Resource:

mix ash_scylla.gen User name:string, email:string

This creates lib/my_app/resources/user.ex with a starter template. Or define it manually:

# lib/my_app/resources/user.ex
defmodule MyApp.User do
use Ash.Resource,
data_layer: AshScylla.DataLayer,
repo: MyApp.Repo
attributes do
uuid_primary_key :id
attribute :name, :string
attribute :email, :string
end
actions do
defaults [:create, :read, :update, :destroy]
end
end

5. Create a Domain:

# lib/my_app/domain.ex
defmodule MyApp.Domain do
use Ash.Domain
resources do
resource MyApp.User
end
end

6. Create Keyspace and Tables:

# Create keyspace (using the mix task)
mix ash_scylla.setup
# Or programmatically
MyApp.Repo.create_keyspace()
# Run migrations
AshScylla.Migrator.run!(MyApp.Repo.nodes(), [
AshScylla.Migration.create_table_cql(MyApp.User),
"CREATE INDEX IF NOT EXISTS idx_users_email ON users (email)"
])

7. Start Using It:

# Create
{:ok, user} = Ash.create(MyApp.User, %{name: "John", email: "john@example.com"})
# Read
users = MyApp.User
|> Ash.Query.filter(email == "john@example.com")
|> Ash.read!()
# Update
{:ok, updated} = user
|> Ash.Changeset.for_update(:update, %{name: "John Doe"})
|> Ash.update()
# Delete
:ok = Ash.destroy(user)

Or using the domain directly:

# Create via domain
{:ok, user} = MyApp.Domain.create_user(%{name: "John", email: "john@example.com"})
# Read via domain
users = MyApp.Domain.read_users!()

Features

Core Ash Features ✅

FeatureStatusDescription
CreateInsert records with TTL support
ReadQuery with filtering and sorting
UpdateUpdate existing records
DestroyDelete records
FilterPowerful filter syntax with CQL WHERE conversion
Sort⚠️ORDER BY on clustering columns only (within a partition)
Keyset paginationToken-based pagination via paging_state (preferred over OFFSET)
LimitLIMIT is natively supported
Offset⚠️Not natively supported in ScyllaDB; results silently truncated. Use keyset pagination instead.
SelectSelect specific fields
MultitenancyKeyspace-based multitenancy
Bulk CreateBatch INSERT operations

ScyllaDB-Specific Features 🚀

TTL (Time To Live)

Automatically expire data after a specified time:

defmodule MyApp.Session do
use Ash.Resource,
data_layer: AshScylla.DataLayer
ash_scylla do
ttl 3600 # Expire after 1 hour
end
end

Consistency Levels

Configure read/write consistency per resource:

ash_scylla do
consistency :quorum # :any, :one, :two, :three, :quorum, :all, :local_quorum
end

Secondary Indexes

Query non-primary key columns efficiently:

ash_scylla do
secondary_index :email # Single column
secondary_index [:name, :age] # Composite index
end

Materialized Views

Create alternative query patterns with automatic view maintenance:

ash_scylla do
materialized_view :users_by_email,
primary_key: [:email, :id],
include_columns: [:name, :age]
end

Batch Operations

Reduce network round-trips with BATCH statements:

# Bulk create (uses BATCH internally)
{:ok, users} = user_data_list
|> Ash.bulk_create(MyApp.User, :create)
# Async partition-aware batching for large datasets
AshScylla.DataLayer.Batch.batch_insert_async(repo, statements, resource: MyApp.User, max_concurrency: 8)

Token-Based Pagination

Efficient pagination without OFFSET:

ash_scylla do
pagination :token # Use token-based pagination instead of OFFSET
end

Per-Action Consistency

Configure consistency levels per action:

ash_scylla do
consistency :quorum # Default consistency
per_action_consistency read: :one, create: :quorum # Per-action overrides
end

Data Modeling Best Practices

ScyllaDB is a wide-column store optimized for specific query patterns. Follow these principles:

1. Query-First Design 🎯

Design your tables around your queries, not the other way around:

# Good: Partition key supports your main query
defmodule MyApp.User do
attributes do
attribute :email, :string, primary_key?: true # Partition key
attribute :name, :string
end
end
# Query by partition key (efficient)
MyApp.User
|> Ash.Query.filter(email == "user@example.com")
|> Ash.read_one()

2. Denormalization is Normal 📦

Duplicate data across tables to support different query patterns:

# Table for querying posts by author
defmodule MyApp.PostByAuthor do
attributes do
attribute :author_id, :uuid, primary_key?: true
attribute :post_id, :uuid, primary_key?: true
attribute :title, :string
attribute :content, :string
end
end
# Table for querying posts by date
defmodule MyApp.PostByDate do
attributes do
attribute :date, :date, primary_key?: true
attribute :post_id, :uuid, primary_key?: true
attribute :title, :string
attribute :author_name, :string # Denormalized
end
end

3. Choose Partition Keys Wisely 🔑

# Good: User ID has high cardinality
attribute :user_id, :uuid, primary_key?: true
# Avoid: Status has low cardinality (creates hotspots)
attribute :status, :string, primary_key?: true # Don't do this

Configuration

Resource Configuration

defmodule MyApp.User do
use Ash.Resource,
data_layer: AshScylla.DataLayer
ash_scylla do
table "users" # Override table name
keyspace "custom_keyspace" # Override keyspace
consistency :quorum # Consistency level
ttl 3600 # Default TTL (seconds)
# Secondary indexes
secondary_index :email
secondary_index [:name, :age]
# Materialized views
materialized_view :users_by_email,
primary_key: [:email, :id],
include_columns: [:name, :age]
end
end

Repo Configuration

config :my_app, MyApp.Repo,
nodes: ["scylla-1:9042", "scylla-2:9042"], # Cluster nodes
keyspace: "my_app_prod",
pool_size: 50, # Connections per node
request_timeout: 300_000, # Query timeout (ms)
connect_timeout: 10_000

Pool Size Guidelines:

ScyllaDB works best with a connections-per-shard approach: pool_size = num_nodes * num_cores_per_node


Limitations

Since ScyllaDB/Cassandra is a NoSQL wide-column store, some features are not supported:

LimitationReasonWorkaround
No JOINsNo relational joinsDenormalize or application-side joins
No complex aggregationsNo GROUP BY, COUNT across partitionsMaterialized views or custom aggregation
No ACID transactionsOnly lightweight transactions (LWT)Use LWT for single-partition operations
Limited WHERE clausesWithout indexes, only PK queries are efficient; filtering on non-indexed columns raises errorsCreate secondary indexes or materialized views for non-PK query patterns
No OR conditionsCQL limitationMultiple queries or UNION-like patterns
No foreign keysNo relational integrityApplication-level validation
OFFSET not supportedScyllaDB has no native OFFSET; it would require full table scanUse keyset pagination with pagination :token. The data layer silently drops OFFSET to prevent performance disasters.

Observability

Telemetry

AshScylla emits standard :telemetry events for all query and batch operations, enabling integration with LiveDashboard, Datadog, OpenTelemetry, and other observability tools.

Query events:

Batch events:

Attaching a handler:

:telemetry.attach(
"ash_scylla-logger",
[:ash_scylla, :query, :stop],
&MyApp.Telemetry.handle_event/4,
nil
)

Prepared Statement Caching

For high-throughput workloads, enable the prepared statement cache to eliminate repeated query parsing overhead on ScyllaDB:

# In your supervision tree
children = [
AshScylla.PreparedStatementCache,
# ... other children
]

Documentation

For detailed documentation, see:


Testing

Run the test suite:

# All tests (unit + integration; requires Podman/Docker for testcontainers)
mix test
# Unit tests only (no ScyllaDB required)
mix test --exclude integration
# Integration tests only (requires Podman/Docker)
mix test test/scylla_integration_test.exs --only integration
# CI pipeline (unit tests + credo)
mix test.ci

Test Structure

FileDescription
test/ash_scylla_test.exsCore DataLayer and DSL unit tests
test/data_layer_crud_test.exsCRUD operations with FakeRepo (create, update, destroy, upsert, bulk_create, run_query, aggregates)
test/data_layer_callbacks_test.exsDataLayer callbacks (transform_query, set_tenant, set_context, filter, sort, limit, offset, select, lock, combination_of, calculate, add_aggregate, add_aggregates, distinct)
test/data_layer_pipeline_test.exsFull pipeline DSL → DataLayer → QueryBuilder → CQL generation and execution
test/data_layer_comprehensive_test.exsComprehensive gap coverage: run_query edge cases, filter OR rewriting, sort edge cases, bulk_create scenarios, source/repo edge cases, upsert delegation, aggregates, distinct, calculate, handle_scylla_result, sanitize_identifier, struct defaults, exhaustive can?/2
test/edge_cases_test.exsEdge cases for QueryBuilder, Batch, Pagination, MaterializedView, Migration
test/error_edge_cases_test.exsComprehensive error handling edge cases
test/dsl_resource_test.exsDSL compilation, public API, secondary_index parsing, materialized_view
test/integration_test.exsIntegration test placeholder
test/scylla_integration_test.exsFull integration tests with testcontainers

Integration tests use testcontainer_ex to spin up a ScyllaDB instance automatically via Podman or Docker.


Contributing

Contributions are welcome! Here's how to get started:

  1. Fork the repository
  2. Clone your fork: git clone https://github.com/your-username/ash_scylla.git
  3. Create a feature branch: git checkout -b feature/my-feature
  4. Make your changes
  5. Run tests: mix test
  6. Commit your changes: git commit -am 'Add some feature'
  7. Push to the branch: git push origin feature/my-feature
  8. Create a Pull Request

Development Setup

# Install dependencies
mix deps.get
# Start ScyllaDB via Podman Compose (includes health checks)
podman-compose up -d
# Or via Docker Compose
docker compose up -d
# Or start ScyllaDB manually with Podman
podman run -p 9042:9042 scylladb/scylla:latest
# Or with Docker
docker run -p 9042:9042 scylladb/scylla:latest
# Run tests
mix test

Dev Container

A .devcontainer/devcontainer.json is provided for VS Code Dev Containers. It brings up both Elixir and ScyllaDB together via Podman Compose or Docker Compose.


License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.


Acknowledgments


Made with ❤️ for the Elixir and Ash communities