Cassandra

What This Skill Does

The Cassandra skill empowers your OpenClaw agent to act as an expert database administrator and architect. It provides deep expertise in Apache Cassandra and DataStax Astra DB ecosystems. The agent is trained to translate complex business requirements into high-performance, denormalized data models, write performant CQL queries, and diagnose performance-killing issues like tombstone accumulation, hot spots, and inefficient partition designs. It ensures your data layer remains scalable, available, and performant by strictly enforcing Cassandra best practices.

Installation

To integrate this skill into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/ivangdavila/cassandra

Use Cases

• Schema Design Review: Analyze existing schemas to ensure partition keys are correctly distributed to avoid hot spots.
• Query Optimization: Transform ALLOW FILTERING queries into efficient, primary-key-based lookups by redesigning the underlying tables.
• Production Troubleshooting: Investigate performance degradation caused by tombstone overhead, mass deletes, or TTL misuse.
• Consistency Tuning: Advise on appropriate consistency levels (e.g., LOCAL_QUORUM vs ONE) for specific read/write workloads in multi-DC setups.
• LWT Management: Architect safe read-modify-write patterns using Lightweight Transactions without sacrificing throughput.

Example Prompts

1. "I am seeing high read latency on my user_activity table. The table uses user_id as a partition key, but some users have millions of rows. How can I optimize this using time-bucketing?"
2. "Explain the performance implications of using a secondary index on a high-cardinality column and suggest a better approach for querying that data."
3. "Help me design a denormalized schema for a retail application where I need to query orders by customer_id, date, and order_status. Cassandra doesn't support JOINs, so what is the best strategy?"

Tips & Limitations

• Denormalization is Mandatory: You must accept data duplication. If you attempt to use Cassandra like a relational database (RDBMS), you will encounter performance bottlenecks.
• Monitor Tombstones: Always keep an eye on nodetool cfstats. If your tombstone count is rising, your query patterns or TTL strategies are likely misconfigured.
• Partition Limits: Keep partitions under 100MB. If a partition grows indefinitely, the node will eventually struggle to manage that slice of data.
• Batching: Never use LOGGED BATCH for general purpose writes; it is only for multi-partition atomicity and carries a heavy performance penalty.
• Avoid ALLOW FILTERING: This is a red flag. If your query requires it, your schema is almost certainly designed incorrectly for the intended access pattern.