Verified Commit a89f6ed3 authored by Marc Vef's avatar Marc Vef
Browse files

Cleanup, Readme, Changelog, Added env variable to LIBGKFS_DIR_CACHE

parent 22e94669

CHANGELOG.md

+3 −0
Original line number Diff line number Diff line
@@ -8,6 +8,9 @@ to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). 
## [Unreleased]

### New



- Added a directory cache for the file system client to improve `ls -l` type operations by avoiding consecutive stat calls

  ([!194](https://storage.bsc.es/gitlab/hpc/gekkofs/-/merge_requests/194)).

  - The cache is experimental and thus disabled by default and can be enabled with the env variable `LIBGKFS_DISABLE_DIR_CACHE` set to `ON`.

- Added file system expansion support ([!196](https://storage.bsc.es/gitlab/hpc/gekkofs/-/merge_requests/196)).

  - Added the tool `gkfs_malleability` to steer start, status, and finalize requests for expansion operations.

  - `-DGKFS_BUILD_TOOLS=ON` must be set for CMake to build the tool.

README.md

+3 −0
Original line number Diff line number Diff line
@@ -517,6 +517,9 @@ Client-metrics require the CMake argument `-DGKFS_ENABLE_CLIENT_METRICS=ON` (see 
- `LIBGKFS_METRICS_IP_PORT` - Enable flushing to a set ZeroMQ server (replaces `LIBGKFS_METRICS_PATH`).

- `LIBGKFS_PROXY_PID_FILE` - Path to the proxy pid file (when using the GekkoFS proxy).

- `LIBGKFS_NUM_REPL` - Number of replicas for data.

#### Caching

- `LIBGKFS_DENTRY_CACHE` - Enable caching directory entries until closing the directory (default: OFF). 

Improves performance for `ls -l` type operations. Further compile-time settings available at `include/config.hpp`.



### Daemon

#### Logging

include/client/cache.hpp

+51 −12
Original line number Diff line number Diff line
@@ -44,25 +44,44 @@ namespace gkfs::cache { 


namespace dir {



/**

 * @brief Cache entry metadata.

 * The entries are limited to the get_dir_extended RPC.

 */

struct cache_entry {

    gkfs::filemap::FileType file_type;

    uint64_t size;

    time_t ctime;

};



/**

 * @brief Cache for directory entries to accelerate ls -l type operations

 */

class DentryCache {

private:

    // <dir_id, <name, cache_entry>>

    // <dir_id, <name, cache_entry>>: Associate a directory id with its entries

    // containing the directory name and cache entry metadata

    std::unordered_map<uint32_t, std::unordered_map<std::string, cache_entry>>

            entries_;

    // <dir_path, dir_id>

    // <dir_path, dir_id>: Associate a directory path with a unique id

    std::unordered_map<std::string, uint32_t> entry_dir_id_;

    std::mutex mtx_;

    std::hash<std::string> str_hash;

    std::mutex mtx_;                 // Mutex to protect the cache

    std::hash<std::string> str_hash; // hash to generate ids



    /**

     * @brief Generate a unique id for caching a directory

     * @param dir_path

     * @return id

     */

    uint32_t

    gen_dir_id(const std::string& dir_path);



    /**

     * @brief Get the unique id for a directory to retrieve its entries. Creates

     * an id if it does not exist.

     * @param dir_path

     * @return id

     */

    uint32_t

    get_dir_id(const std::string& dir_path);
@@ -71,28 +90,48 @@ public: 


    virtual ~DentryCache() = default;



    /**

     * @brief Insert a new entry in the cache

     * @param parent_dir

     * @param name

     * @param value

     */

    void

    insert(const std::string& parent_dir, const std::string name,

           const cache_entry value);

    insert(const std::string& parent_dir, std::string name, cache_entry value);



    /**

     * @brief Get an entry from the cache for a given directory

     * @param parent_dir

     * @param name

     * @return std::optional<cache_entry>

     */

    std::optional<cache_entry>

    get(const std::string& parent_dir, const std::string& name);



    /**

     * @brief Clear the cache for a given directory. Called when a directory is

     * closed

     * @param dir_path

     */

    void

    clear_dir(const std::string& dir_path);



    /**

     * @brief Dump the cache to the log for debugging purposes. Not used in

     * production.

     * @param dir_path

     */

    void

    dump_cache_to_log(const std::string& dir_path);



    /**

     * @brief Clear the entire cache

     */

    void

    clear();

};

} // namespace dir



//// <path<cnt, size>>

// std::unordered_map<std::string, std::pair<size_t, size_t>> size_cache;





} // namespace gkfs::cache



#endif // GKFS_CLIENT_CACHE

include/client/env.hpp

+1 −1
Original line number Diff line number Diff line
@@ -60,7 +60,7 @@ static constexpr auto METRICS_IP_PORT = ADD_PREFIX("METRICS_IP_PORT"); 


static constexpr auto NUM_REPL = ADD_PREFIX("NUM_REPL");

static constexpr auto PROXY_PID_FILE = ADD_PREFIX("PROXY_PID_FILE");

static constexpr auto DISABLE_DENTRY_CACHE = ADD_PREFIX("DISABLE_DENTRY_CACHE");

static constexpr auto DENTRY_CACHE = ADD_PREFIX("DENTRY_CACHE");



} // namespace gkfs::env

include/client/preload_context.hpp

+1 −4
Original line number Diff line number Diff line
@@ -59,10 +59,7 @@ namespace cache { 
namespace dir {

class DentryCache;

}

namespace file {

class WriteSizeCache;

}

}

} // namespace cache



namespace preload {

/*