// Licensed to the Apache Software Foundation (ASF) under one

// or more contributor license agreements.  See the NOTICE file

// distributed with this work for additional information

// regarding copyright ownership.  The ASF licenses this file

// to you under the Apache License, Version 2.0 (the

// "License"); you may not use this file except in compliance

// with the License.  You may obtain a copy of the License at

//

//   http://www.apache.org/licenses/LICENSE-2.0

//

// Unless required by applicable law or agreed to in writing,

// software distributed under the License is distributed on an

// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY

// KIND, either express or implied.  See the License for the

// specific language governing permissions and limitations

// under the License.

#pragma once

#include <bvar/bvar.h>

#include <concurrentqueue.h>

#include <algorithm>

#include <array>

#include <atomic>

#include <boost/lockfree/spsc_queue.hpp>

#include <condition_variable>

#include <functional>

#include <limits>

#include <map>

#include <memory>

#include <mutex>

#include <optional>

#include <thread>

#include <unordered_map>

#include <unordered_set>

#include <vector>

#include "io/cache/block_file_cache_ttl_mgr.h"

#include "io/cache/cache_lru_dumper.h"

#include "io/cache/file_block.h"

#include "io/cache/file_cache_common.h"

#include "io/cache/file_cache_storage.h"

#include "io/cache/lru_queue_recorder.h"

#include "runtime/runtime_profile.h"

namespace doris::io {

using RecycleFileCacheKeys = moodycamel::ConcurrentQueue<FileCacheKey>;

class LockScopedTimer {

public:

    LockScopedTimer() : start_(std::chrono::steady_clock::now()) {}

    ~LockScopedTimer() {

        auto end = std::chrono::steady_clock::now();

        auto duration_us =

                std::chrono::duration_cast<std::chrono::microseconds>(end - start_).count();

        if (duration_us > config::cache_lock_held_long_tail_threshold_us) {

            LOG(WARNING) << "Lock held time " << std::to_string(duration_us) << "us. "

                         << get_stack_trace();

        }

    }

private:

    std::chrono::time_point<std::chrono::steady_clock> start_;

};

// Note: the cache_lock is scoped, so do not add do...while(0) here.

#define SCOPED_CACHE_LOCK(MUTEX, cache)                                                           \

    std::chrono::time_point<std::chrono::steady_clock> start_time =                               \

            std::chrono::steady_clock::now();                                                     \

    std::lock_guard cache_lock(MUTEX);                                                            \

    std::chrono::time_point<std::chrono::steady_clock> acq_time =                                 \

            std::chrono::steady_clock::now();                                                     \

    auto duration_us =                                                                            \

            std::chrono::duration_cast<std::chrono::microseconds>(acq_time - start_time).count(); \

    *(cache->_cache_lock_wait_time_us) << duration_us;                                            \

    if (duration_us > config::cache_lock_wait_long_tail_threshold_us) {                           \

        LOG(WARNING) << "Lock wait time " << std::to_string(duration_us) << "us. "                \

                     << get_stack_trace() << std::endl;                                           \

    }                                                                                             \

    LockScopedTimer cache_lock_timer;

class AsyncCacheWriteManager;

class FSFileCacheStorage;

class InflightWriteBufferIndex;

struct FileBlocksProbeResult {

    explicit FileBlocksProbeResult(std::vector<FileBlockSPtr> file_blocks_)

            : file_blocks(std::move(file_blocks_)) {}

    FileBlocksProbeResult(FileBlocksProbeResult&&) noexcept = default;

    FileBlocksProbeResult& operator=(FileBlocksProbeResult&&) noexcept = delete;

    FileBlocksProbeResult(const FileBlocksProbeResult&) = delete;

    FileBlocksProbeResult& operator=(const FileBlocksProbeResult&) = delete;

    ~FileBlocksProbeResult();

    /// One entry per cache-block-sized input slot, in offset order. A null entry is a cache miss;

    /// a non-null entry covers the whole slot. Its right boundary can exceed the final short slot

    /// while a file writer still owns a full-size preallocated tail block. Retaining and releasing

    /// a probe result never acquires or completes downloader ownership.

    std::vector<FileBlockSPtr> file_blocks;

};

// NeedUpdateLRUBlocks keeps FileBlockSPtr entries that require LRU updates in a

// deduplicated, sharded container. Entries are keyed by the raw FileBlock

// pointer so that multiple shared_ptr copies of the same block are treated as a

// single pending update. The structure is thread-safe and optimized for high

// contention insert/drain workloads in the background update thread.

// Note that Blocks are updated in batch, internal order is not important.

class NeedUpdateLRUBlocks {

public:

    NeedUpdateLRUBlocks() = default;

    // Insert a block into the pending set. Returns true only when the block

    // was not already queued. Null inputs are ignored.

    bool insert(FileBlockSPtr block, size_t max_queue_size = std::numeric_limits<size_t>::max());

    // Drain up to `limit` unique blocks into `output`. The method returns how

    // many blocks were actually drained and shrinks the internal size

    // accordingly.

    size_t drain(size_t limit, std::vector<FileBlockSPtr>* output);

    // Remove every pending block from the structure and reset the size.

    void clear();

    // Thread-safe approximate size of queued unique blocks.

    size_t size() const { return _size.load(std::memory_order_relaxed); }

private:

    static constexpr size_t kShardCount = 64;

    static constexpr size_t kShardMask = kShardCount - 1;

    struct Shard {

        std::mutex mutex;

        std::unordered_map<FileBlock*, FileBlockSPtr> entries;

    };

    size_t shard_index(FileBlock* ptr) const;

    void decrease_size(size_t delta);

    std::array<Shard, kShardCount> _shards;

    std::atomic<size_t> _size {0};

};

// The BlockFileCache is responsible for the management of the blocks

// The current strategies are lru and ttl.

struct FileBlockCell {

    friend class FileBlock;

    FileBlockSPtr file_block;

    /// Iterator is put here on first reservation attempt, if successful.

    std::optional<LRUQueue::Iterator> queue_iterator;

    mutable int64_t atime {0};

    void update_atime() const {

        atime = std::chrono::duration_cast<std::chrono::seconds>(

                        std::chrono::steady_clock::now().time_since_epoch())

                        .count();

    }

    /// Pointer to file block is always hold by the cache itself. Every other strong reference

    /// means somebody is still using the block. CachedRemoteFileReader keeps only weak references

    /// in its direct-read map, so however many readers know about a block, they do not keep it

    /// busy.

    bool releasable() const { return file_block.use_count() == 1; }

    size_t size() const { return file_block->_block_range.size(); }

    FileBlockCell() = default;

    FileBlockCell(FileBlockSPtr file_block, std::lock_guard<std::mutex>& cache_lock);

    FileBlockCell(FileBlockCell&& other) noexcept

            : file_block(std::move(other.file_block)),

              queue_iterator(other.queue_iterator),

              atime(other.atime) {

        file_block->cell = this;

    }

    FileBlockCell& operator=(const FileBlockCell&) = delete;

    FileBlockCell(const FileBlockCell&) = delete;

    size_t dowloading_size() const { return file_block->_downloaded_size; }

};

class BlockFileCache {

    friend class FSFileCacheStorage;

    friend class MemFileCacheStorage;

    friend class FileBlock;

    friend struct FileBlocksHolder;

    friend class CacheLRUDumper;

    friend class LRUQueueRecorder;

    friend struct FileBlockCell;

    friend class BlockFileCacheTest;

public:

    // hash the file_name to uint128

    static UInt128Wrapper hash(const std::string& path);

    BlockFileCache(const std::string& cache_base_path, const FileCacheSettings& cache_settings);

    virtual ~BlockFileCache();

    /// Restore cache from local filesystem.

    Status initialize();

    /// Cache capacity in bytes.

    [[nodiscard]] size_t capacity() const { return _capacity; }

    // try to release all releasable block

    // it maybe hang the io/system

    size_t try_release();

    [[nodiscard]] const std::string& get_base_path() const { return _cache_base_path; }

    // Get storage for inspection

    FileCacheStorage* get_storage() const { return _storage.get(); }

    /**

         * Given an `offset` and `size` representing [offset, offset + size) bytes interval,

         * return list of cached non-overlapping non-empty

         * file blocks `[block1, ..., blockN]` which intersect with given interval.

         *

         * blocks in returned list are ordered in ascending order and represent a full contiguous

         * interval (no holes). Each block in returned list has state: DOWNLOADED, DOWNLOADING or EMPTY.

         *

         * As long as pointers to returned file blocks are hold

         * it is guaranteed that these file blocks are not removed from cache.

         */

    FileBlocksHolder get_or_set(const UInt128Wrapper& hash, size_t offset, size_t size,

                                CacheContext& context);

    /// Whether existing DOWNLOADED/DOWNLOADING blocks fully cover `[offset, offset + size)`.

    /// Handles arbitrary block boundaries, including restored upload blocks. May load metadata,

    /// but never reserves cache cells, waits for downloads or touches LRU state.

    bool is_range_downloaded_or_downloading(const UInt128Wrapper& hash, size_t offset, size_t size,

                                            const CacheContext& context);

    /// Probe the block-aligned `[offset, offset + size)` range without creating cache cells or

    /// touching LRU state. The result contains one ordered slot per cache block; each slot is null

    /// on miss or owns an existing block that starts at and covers the slot. A final short slot can

    /// be covered by a full-size block preallocated by a file writer. `context` supplies cache

    /// metadata when lazy loading is required.

    FileBlocksProbeResult probe(const UInt128Wrapper& hash, size_t offset, size_t size,

                                const CacheContext& context);

    /// Touch `block` after a successful local read, provided it is still the cached cell. The

    /// supplied `context` controls the target LRU queue and query-level accounting.

    void touch_probe_block_if_cached(const FileBlockSPtr& block, const CacheContext& context);

    /// Check whether `block` is being deleted while taking cache/block locks in canonical order.

    bool is_block_deleting(const FileBlockSPtr& block) const;

    /// Return this cache disk's async-write manager.

    AsyncCacheWriteManager* async_write_manager() const { return _async_write_manager.get(); }

    /// Return this cache disk's inflight payload index.

    InflightWriteBufferIndex* inflight_write_buffer_index() const {

        return _inflight_write_buffer_index.get();

    }

    /**

     * Return existing downloaded blocks only if they fully cover [offset, offset + size).

     * This lookup is read-only: it does not reserve cache space or create EMPTY blocks.

     */

    Status get_downloaded_blocks_if_fully_covered(const UInt128Wrapper& hash, size_t offset,

                                                  size_t size, const CacheContext& context,

                                                  FileBlocks* blocks, bool* fully_covered);

    /// Diagnostic snapshot of resident DOWNLOADED intervals, clipped to the requested range.

    /// Handles irregular block boundaries; never loads metadata, reserves space or touches LRU.

    std::vector<FileBlock::Range> downloaded_ranges(const UInt128Wrapper& hash, size_t offset,

                                                    size_t size);

    /**

     * record blocks read directly by CachedRemoteFileReader

     */

    void add_need_update_lru_block(FileBlockSPtr block);

    /**

     * Clear all cached data for this cache instance async

     *

     * @returns summary message

     */

    std::string clear_file_cache_async();

    std::string clear_file_cache_sync();

    /**

     * Reset the cache capacity. If the new_capacity is smaller than _capacity, the redundant data will be remove async.

     *

     * @returns summary message

     */

    std::string reset_capacity(size_t new_capacity);

    std::map<size_t, FileBlockSPtr> get_blocks_by_key(const UInt128Wrapper& hash);

    /// For debug and UT

    std::string dump_structure(const UInt128Wrapper& hash);

    std::string dump_single_cache_type(const UInt128Wrapper& hash, size_t offset);

    void dump_lru_queues(bool force);

    [[nodiscard]] size_t get_used_cache_size(FileCacheType type) const;

    [[nodiscard]] size_t get_file_blocks_num(FileCacheType type) const;

    // change the block cache type

    void change_cache_type(const UInt128Wrapper& hash, size_t offset, FileCacheType new_type,

                           std::lock_guard<std::mutex>& cache_lock);

    // remove all blocks that belong to the key

    void remove_if_cached(const UInt128Wrapper& key);

    void remove_if_cached_async(const UInt128Wrapper& key);

    // Reset the block size and keep FileBlock, LRU queue, and cache counters consistent.

    void reset_range(const UInt128Wrapper&, size_t offset, size_t old_size, size_t new_size,

                     std::lock_guard<std::mutex>& cache_lock);

    // get the hotest blocks message by key

    // The tuple is composed of <offset, size, cache_type, expiration_time>

    [[nodiscard]] std::vector<std::tuple<size_t, size_t, FileCacheType, uint64_t>>

    get_hot_blocks_meta(const UInt128Wrapper& hash) const;

    [[nodiscard]] bool get_async_open_success() const { return _async_open_done; }

    // The manager that keeps cached blocks in the cache type their tablet's TTL asks for.

    // Exposed so that tests can drive it deterministically rather than race its threads.

    BlockFileCacheTtlMgr* get_ttl_mgr() { return _ttl_mgr.get(); }

    BlockFileCache& operator=(const BlockFileCache&) = delete;

    BlockFileCache(const BlockFileCache&) = delete;

    // try to reserve the new space for the new block if the cache is full

    bool try_reserve(const UInt128Wrapper& hash, const CacheContext& context, size_t offset,

                     size_t size, std::lock_guard<std::mutex>& cache_lock);

    /**

     * Proactively evict cache blocks to free up space before cache is full.

     * 

     * This function attempts to evict blocks from both NORMAL and TTL queues to maintain 

     * cache size below high watermark. Unlike try_reserve() which blocks until space is freed,

     * this function initiates asynchronous eviction in background.

     * 

     * @param size Number of bytes to try to evict

     * @param cache_lock Lock that must be held while accessing cache data structures

     * 

     * @pre Caller must hold cache_lock

     * @pre _need_evict_cache_in_advance must be true

     * @pre _recycle_keys queue must have capacity for evicted blocks

     */

    void try_evict_in_advance(size_t size, std::lock_guard<std::mutex>& cache_lock);

    void update_ttl_atime(const UInt128Wrapper& hash);

    std::map<std::string, double> get_stats();

    // for be UTs

    std::map<std::string, double> get_stats_unsafe();

    [[nodiscard]] size_t need_update_lru_blocks_size_unsafe() const {

        return _need_update_lru_blocks.size();

    }

    using AccessRecord =

            std::unordered_map<AccessKeyAndOffset, LRUQueue::Iterator, KeyAndOffsetHash>;

    /// Used to track and control the cache access of each query.

    /// Through it, we can realize the processing of different queries by the cache layer.

    struct QueryFileCacheContext {

        LRUQueue lru_queue;

        AccessRecord records;

        QueryFileCacheContext(size_t max_cache_size) : lru_queue(max_cache_size, 0, 0) {}

        void remove(const UInt128Wrapper& hash, size_t offset,

                    std::lock_guard<std::mutex>& cache_lock);

        void reserve(const UInt128Wrapper& hash, size_t offset, size_t size,

                     std::lock_guard<std::mutex>& cache_lock);

        size_t get_max_cache_size() const { return lru_queue.get_max_size(); }

        size_t get_cache_size(std::lock_guard<std::mutex>& cache_lock) const {

            return lru_queue.get_capacity(cache_lock);

        }

        LRUQueue& queue() { return lru_queue; }

    };

    using QueryFileCacheContextPtr = std::shared_ptr<QueryFileCacheContext>;

    using QueryFileCacheContextMap = std::unordered_map<TUniqueId, QueryFileCacheContextPtr>;

    QueryFileCacheContextPtr get_query_context(const TUniqueId& query_id,

                                               std::lock_guard<std::mutex>&);

    void remove_query_context(const TUniqueId& query_id);

    QueryFileCacheContextPtr get_or_set_query_context(const TUniqueId& query_id,

                                                      std::lock_guard<std::mutex>& cache_lock,

                                                      int file_cache_query_limit_percent);

    /// Save a query context information, and adopt different cache policies

    /// for different queries through the context cache layer.

    struct QueryFileCacheContextHolder {

        QueryFileCacheContextHolder(const TUniqueId& query_id, BlockFileCache* mgr,

                                    QueryFileCacheContextPtr context)

                : query_id(query_id), mgr(mgr), context(context) {}

        QueryFileCacheContextHolder& operator=(const QueryFileCacheContextHolder&) = delete;

        QueryFileCacheContextHolder(const QueryFileCacheContextHolder&) = delete;

        ~QueryFileCacheContextHolder() {

            /// If only the query_map and the current holder hold the context_query,

            /// the query has been completed and the query_context is released.

            if (context) {

                context.reset();

                mgr->remove_query_context(query_id);

            }

        }

        const TUniqueId& query_id;

        BlockFileCache* mgr = nullptr;

        QueryFileCacheContextPtr context;

    };

    using QueryFileCacheContextHolderPtr = std::unique_ptr<QueryFileCacheContextHolder>;

    QueryFileCacheContextHolderPtr get_query_context_holder(const TUniqueId& query_id,

                                                            int file_cache_query_limit_percent);

    int64_t approximate_available_cache_size() const {

        return std::max<int64_t>(

                _cache_capacity_metrics->get_value() - _cur_cache_size_metrics->get_value(), 0);

    }

    Status report_file_cache_inconsistency(std::vector<std::string>& results);

    Status check_file_cache_consistency(InconsistencyContext& inconsistency_context);

private:

    // Shared scan used by both clear modes. It keeps the FileBlock holder lifecycle intact:

    // releasable blocks are removed immediately, while blocks held by readers are only marked

    // deleting and are later removed by FileBlocksHolder destruction or, when the last holder

    // does not release through FileBlock::release_cache_reference(), by run_background_gc().

    std::string clear_file_cache_impl(bool sync_remove);

    LRUQueue& get_queue(FileCacheType type);

    const LRUQueue& get_queue(FileCacheType type) const;

    template <class T, class U>

        requires IsXLock<T> && IsXLock<U>

    void remove(FileBlockSPtr file_block, T& cache_lock, U& segment_lock, bool sync = true);

    FileBlocks get_impl(const UInt128Wrapper& hash, const CacheContext& context,

                        const FileBlock::Range& range, std::lock_guard<std::mutex>& cache_lock);

    template <class T>

        requires IsXLock<T>

    FileBlockCell* get_cell(const UInt128Wrapper& hash, size_t offset, T& cache_lock);

    virtual FileBlockCell* add_cell(const UInt128Wrapper& hash, const CacheContext& context,

                                    size_t offset, size_t size, FileBlock::State state,

                                    std::lock_guard<std::mutex>& cache_lock);

    Status initialize_unlocked(std::lock_guard<std::mutex>& cache_lock);

    void update_block_lru(FileBlockSPtr block, std::lock_guard<std::mutex>& cache_lock);

    void use_cell(const FileBlockCell& cell, FileBlocks* result, bool not_need_move,

                  std::lock_guard<std::mutex>& cache_lock);

    bool try_reserve_for_lru(const UInt128Wrapper& hash, QueryFileCacheContextPtr query_context,

                             const CacheContext& context, size_t offset, size_t size,

                             std::lock_guard<std::mutex>& cache_lock,

                             bool evict_in_advance = false);

    bool try_reserve_during_async_load(size_t size, std::lock_guard<std::mutex>& cache_lock);

    std::vector<FileCacheType> get_other_cache_type(FileCacheType cur_cache_type);

    std::vector<FileCacheType> get_other_cache_type_without_ttl(FileCacheType cur_cache_type);

    bool try_reserve_from_other_queue(FileCacheType cur_cache_type, size_t offset, int64_t cur_time,

                                      std::lock_guard<std::mutex>& cache_lock,

                                      bool evict_in_advance = false);

    size_t get_available_cache_size(FileCacheType cache_type) const;

    FileBlocks split_range_into_cells(const UInt128Wrapper& hash, const CacheContext& context,

                                      size_t offset, size_t size, FileBlock::State state,

                                      std::lock_guard<std::mutex>& cache_lock);

    std::string dump_structure_unlocked(const UInt128Wrapper& hash,

                                        std::lock_guard<std::mutex>& cache_lock);

    std::string dump_single_cache_type_unlocked(const UInt128Wrapper& hash, size_t offset,

                                                std::lock_guard<std::mutex>& cache_lock);

    void fill_holes_with_empty_file_blocks(FileBlocks& file_blocks, const UInt128Wrapper& hash,

                                           const CacheContext& context,

                                           const FileBlock::Range& range,

                                           std::lock_guard<std::mutex>& cache_lock);

    size_t get_used_cache_size_unlocked(FileCacheType type,

                                        std::lock_guard<std::mutex>& cache_lock) const;

    void check_disk_resource_limit();

    void check_need_evict_cache_in_advance();

    size_t get_available_cache_size_unlocked(FileCacheType type,

                                             std::lock_guard<std::mutex>& cache_lock) const;

    size_t get_file_blocks_num_unlocked(FileCacheType type,

                                        std::lock_guard<std::mutex>& cache_lock) const;

    bool need_to_move(FileCacheType cell_type, FileCacheType query_type) const;

    void run_background_monitor();

    void run_background_gc();

    void run_background_lru_log_replay();

    size_t replay_lru_logs_once();

    void run_background_lru_dump();

    void restore_lru_queues_from_disk(std::lock_guard<std::mutex>& cache_lock);

    void run_background_evict_in_advance();

    void run_background_block_lru_update();

    bool try_reserve_from_other_queue_by_time_interval(FileCacheType cur_type,

                                                       std::vector<FileCacheType> other_cache_types,

                                                       size_t size, int64_t cur_time,

                                                       std::lock_guard<std::mutex>& cache_lock,

                                                       bool evict_in_advance);

    bool try_reserve_from_other_queue_by_size(FileCacheType cur_type,

                                              std::vector<FileCacheType> other_cache_types,

                                              size_t size, std::lock_guard<std::mutex>& cache_lock,

                                              bool evict_in_advance);

    bool is_overflow(size_t removed_size, size_t need_size, size_t cur_cache_size,

                     bool evict_in_advance) const;

    void remove_file_blocks(std::vector<FileBlockCell*>&, std::lock_guard<std::mutex>&, bool sync,

                            std::string& reason);

    void find_evict_candidates(LRUQueue& queue, size_t size, size_t cur_cache_size,

                               size_t& removed_size, std::vector<FileBlockCell*>& to_evict,

                               std::lock_guard<std::mutex>& cache_lock, size_t& cur_removed_size,

                               bool evict_in_advance);

    Status check_ofstream_status(std::ofstream& out, std::string& filename);

    Status dump_one_lru_entry(std::ofstream& out, std::string& filename, const UInt128Wrapper& hash,

                              size_t offset, size_t size);

    Status finalize_dump(std::ofstream& out, size_t entry_num, std::string& tmp_filename,

                         std::string& final_filename, size_t& file_size);

    Status check_ifstream_status(std::ifstream& in, std::string& filename);

    Status parse_dump_footer(std::ifstream& in, std::string& filename, size_t& entry_num);

    Status parse_one_lru_entry(std::ifstream& in, std::string& filename, UInt128Wrapper& hash,

                               size_t& offset, size_t& size);

    void remove_lru_dump_files();

    void clear_need_update_lru_blocks();

    // Marks a cell that cannot be erased yet as deleting and remembers it in _deleting_blocks, so

    // that the block is still reclaimed when its last reference is dropped without going through

    // FileBlock::release_cache_reference(). This is the only place that sets a block deleting.

    template <class T>

        requires IsXLock<T>

    void mark_cell_deleting(FileBlockCell& cell, T& cache_lock);

    // Removes up to batch_limit blocks that were marked deleting and are not referenced any

    // more. Returns how many of them were removed.

    size_t recycle_deleting_blocks(size_t batch_limit);

    // info

    std::string _cache_base_path;

    size_t _capacity = 0;

    size_t _max_file_block_size = 0;

    mutable std::mutex _mutex;

    bool _close {false};

    std::mutex _close_mtx;

    std::condition_variable _close_cv;

    std::thread _cache_background_monitor_thread;

    std::thread _cache_background_gc_thread;

    std::thread _cache_background_evict_in_advance_thread;

    std::thread _cache_background_lru_dump_thread;

    std::thread _cache_background_lru_log_replay_thread;

    std::thread _cache_background_block_lru_update_thread;

    std::atomic_bool _async_open_done {false};

    // disk space or inode is less than the specified value

    std::atomic<bool> _disk_resource_limit_mode {false};

    std::atomic<bool> _need_evict_cache_in_advance {false};

    bool _is_initialized {false};

    // strategy

    using FileBlocksByOffset = std::map<size_t, FileBlockCell>;

    using CachedFiles = std::unordered_map<UInt128Wrapper, FileBlocksByOffset, KeyHash>;

    CachedFiles _files;

    QueryFileCacheContextMap _query_map;

    size_t _cur_cache_size = 0;

    size_t _cur_ttl_size = 0;

    std::multimap<uint64_t, UInt128Wrapper> _time_to_key;

    std::unordered_map<UInt128Wrapper, uint64_t, KeyHash> _key_to_time;

    // The three queues are level queue.

    // It means as level1/level2/level3 queue.

    // but the level2 is maximum.

    // If some datas are importance, we can cache it into index queue

    // If some datas are just use once, we can cache it into disposable queue

    // The size proportion is [1:17:2].

    LRUQueue _index_queue;

    LRUQueue _normal_queue;

    LRUQueue _disposable_queue;

    LRUQueue _ttl_queue;

    // keys for async remove

    RecycleFileCacheKeys _recycle_keys;

    // Blocks that were busy when something asked to delete them. They keep sitting in _files

    // until their last reference goes away. Only FileBlocksHolder and FileBlocksProbeResult

    // release through FileBlock::release_cache_reference(), which is the one path that notices

    // is_deleting and removes the block; every other reference holder just drops its shared_ptr

    // and would strand the block here forever. run_background_gc() rechecks these candidates and

    // removes the ones that became releasable.

    //

    // Invariant: holds exactly the keys of the cells in _files whose block is deleting. A key is

    // added by mark_cell_deleting() and erased by remove() together with its cell, so a block

    // created later at the same offset is never taken for the one that was marked.

    std::unordered_set<AccessKeyAndOffset, KeyAndOffsetHash> _deleting_blocks;

    std::unique_ptr<LRUQueueRecorder> _lru_recorder;

    std::unique_ptr<CacheLRUDumper> _lru_dumper;

    std::unique_ptr<BlockFileCacheTtlMgr> _ttl_mgr;

    std::unique_ptr<InflightWriteBufferIndex> _inflight_write_buffer_index;

    std::unique_ptr<AsyncCacheWriteManager> _async_write_manager;

    // metrics

    std::shared_ptr<bvar::Status<size_t>> _cache_capacity_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_cache_size_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_ttl_cache_size_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_ttl_cache_lru_queue_cache_size_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_ttl_cache_lru_queue_element_count_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_normal_queue_element_count_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_normal_queue_cache_size_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_index_queue_element_count_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_index_queue_cache_size_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_disposable_queue_element_count_metrics;

    std::shared_ptr<bvar::Status<size_t>> _cur_disposable_queue_cache_size_metrics;

    std::array<std::shared_ptr<bvar::Adder<size_t>>, 4> _queue_evict_size_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _total_read_size_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _total_hit_size_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _total_evict_size_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _evict_not_downloaded_size_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _evict_not_downloaded_num_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _gc_evict_bytes_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _gc_evict_count_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _evict_by_time_metrics_matrix[4][4];

    std::shared_ptr<bvar::Adder<size_t>> _evict_by_size_metrics_matrix[4][4];

    std::shared_ptr<bvar::Adder<size_t>> _evict_by_self_lru_metrics_matrix[4];

    std::shared_ptr<bvar::Adder<size_t>> _evict_by_try_release;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _num_hit_blocks_5m;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _num_read_blocks_5m;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _num_hit_blocks_1h;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _num_read_blocks_1h;

    std::shared_ptr<bvar::Adder<size_t>> _num_read_blocks;

    std::shared_ptr<bvar::Adder<size_t>> _num_hit_blocks;

    std::shared_ptr<bvar::Adder<size_t>> _num_removed_blocks;

    std::shared_ptr<bvar::Adder<size_t>> _no_warmup_num_read_blocks;

    std::shared_ptr<bvar::Adder<size_t>> _no_warmup_num_hit_blocks;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _no_warmup_num_hit_blocks_5m;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _no_warmup_num_read_blocks_5m;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _no_warmup_num_hit_blocks_1h;

    std::shared_ptr<bvar::Window<bvar::Adder<size_t>>> _no_warmup_num_read_blocks_1h;

    std::shared_ptr<bvar::Status<double>> _hit_ratio;

    std::shared_ptr<bvar::Status<double>> _hit_ratio_5m;

    std::shared_ptr<bvar::Status<double>> _hit_ratio_1h;

    std::shared_ptr<bvar::Status<double>> _no_warmup_hit_ratio;

    std::shared_ptr<bvar::Status<double>> _no_warmup_hit_ratio_5m;

    std::shared_ptr<bvar::Status<double>> _no_warmup_hit_ratio_1h;

    std::shared_ptr<bvar::Status<size_t>> _disk_limit_mode_metrics;

    std::shared_ptr<bvar::Status<size_t>> _need_evict_cache_in_advance_metrics;

    std::shared_ptr<bvar::Status<size_t>> _meta_store_write_queue_size_metrics;

    std::shared_ptr<bvar::LatencyRecorder> _cache_lock_wait_time_us;

    std::shared_ptr<bvar::LatencyRecorder> _get_or_set_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _probe_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _storage_sync_remove_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _storage_retry_sync_remove_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _storage_async_remove_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _evict_in_advance_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _recycle_keys_length_recorder;

    std::shared_ptr<bvar::Status<size_t>> _deleting_block_count_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _recycle_deleting_blocks_metrics;

    std::shared_ptr<bvar::LatencyRecorder> _update_lru_blocks_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _need_update_lru_blocks_length_recorder;

    std::shared_ptr<bvar::Adder<size_t>> _need_update_lru_blocks_produce_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _need_update_lru_blocks_consume_metrics;

    std::shared_ptr<bvar::LatencyRecorder> _ttl_gc_latency_us;

    std::shared_ptr<bvar::LatencyRecorder> _shadow_queue_levenshtein_distance;

    std::array<std::shared_ptr<bvar::LatencyRecorder>, 4> _lru_recorder_queue_length_recorder;

    std::array<std::shared_ptr<bvar::Adder<size_t>>, 4> _lru_recorder_queue_produce_metrics;

    std::array<std::shared_ptr<bvar::Adder<size_t>>, 4> _lru_recorder_queue_consume_metrics;

    std::array<std::shared_ptr<bvar::Status<size_t>>, 4>

            _lru_recorder_shadow_queue_element_count_metrics;

    std::shared_ptr<bvar::Adder<size_t>> _lru_recorder_log_replay_idle_metrics;

    // keep _storage last so it will deconstruct first

    // otherwise, load_cache_info_into_memory might crash

    // coz it will use other members of BlockFileCache

    // so join this async load thread first

    std::unique_ptr<FileCacheStorage> _storage;

    std::shared_ptr<bvar::LatencyRecorder> _lru_dump_latency_us;

    std::mutex _dump_lru_queues_mtx;

    NeedUpdateLRUBlocks _need_update_lru_blocks;

};

} // namespace doris::io