The Complete Production SOP: How to Deploy and Run RocksDB LSM-Tree Write Optimization & Compaction in Autonomous AI Agents

---
name: "rocksdb-lsm-tree-compaction-tuning"
description: "Tune RocksDB embedded key-value storage engine for extreme write throughput and SSD endurance."
metadata:
  author: "Meta"
  version: "1.0.0"
  runtime_compatibility:
    - "codex"
    - "claude-code"
    - "cursor"
    - "windsurf"
triggers:
  - "tune rocksdb compaction"
  - "configure rocksdb block cache"
  - "optimize rocksdb write throughput"
permissions:
  fs_read: true
  fs_write: "workspace"
  shell_exec: false
  net_access: false
---

# RocksDB LSM-Tree Write Optimization & Compaction

## Execution SOP & Rules
1. Set options.write_buffer_size = 64 * 1024 * 1024 (64MB) and max_write_buffer_number = 4.
2. Configure BlockBasedTableOptions with NewBloomFilterPolicy(10, false) for fast key existence checks.
3. Allocate shared block cache: table_options.block_cache = NewLRUCache(16 * 1024 * 1024 * 1024ULL).
4. Pin index and filter blocks into cache: table_options.cache_index_and_filter_blocks = true.
5. Monitor performance statistics using DB::GetProperty('rocksdb.stats', &stats).

#pragma once
#include <rocksdb/db.h>
#include <rocksdb/table.h>
#include <rocksdb/filter_policy.h>
#include <rocksdb/cache.h>

inline rocksdb::Options GetOptimizedRocksDBOptions(std::shared_ptr<rocksdb::Cache> cache) {
    rocksdb::Options options;
    options.create_if_missing = true;
    options.write_buffer_size = 64 * 1024 * 1024;
    options.max_write_buffer_number = 4;
    options.min_write_buffer_number_to_merge = 2;
    options.level0_file_num_compaction_trigger = 4;
    options.level0_slowdown_writes_trigger = 20;
    options.level0_stop_writes_trigger = 36;
    options.max_background_jobs = 6;

    rocksdb::BlockBasedTableOptions table_options;
    table_options.block_cache = cache;
    table_options.filter_policy.reset(rocksdb::NewBloomFilterPolicy(10, false));
    table_options.cache_index_and_filter_blocks = true;
    table_options.pin_l0_filter_and_index_blocks_in_cache = true;
    options.table_factory.reset(rocksdb::NewBlockBasedTableFactory(table_options));
    return options;
}

# Execute via Claude Code CLI to confirm skill triggering
claude --print "optimize rocksdb write throughput for high-ingestion telemetry store"

#include "storage_options.hpp"
#include <cassert>
#include <iostream>

int main() {
    auto cache = rocksdb::NewLRUCache(1ULL * 1024 * 1024 * 1024);
    rocksdb::Options options = GetOptimizedRocksDBOptions(cache);
    rocksdb::DB* db = nullptr;
    rocksdb::Status s = rocksdb::DB::Open(options, "/tmp/rocksdb_test_bench", &db);
    assert(s.ok());

    std::string stats;
    db->GetProperty("rocksdb.stats", &stats);
    std::cout << "RocksDB Options Verified. Engine Stats:\n" << stats << std::endl;
    delete db;
    return 0;
}