Limit trash directory to be 25% of total DB

Summary:
Update DeleteScheduler to delete files immediately if trash directory is >= 25% of DB size
Closes https://github.com/facebook/rocksdb/pull/2436

Differential Revision: D5230384

Pulled By: IslamAbdelRahman

fbshipit-source-id: 5cbda8ac536a3cc72c774641621edc02c8202482

db/db_sst_test.cc

@@ -337,6 +337,7 @@ TEST_F(DBSSTTest, RateLimitedDelete) {
   ASSERT_OK(s);
   options.sst_file_manager->SetDeleteRateBytesPerSecond(rate_bytes_per_sec);
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+  sfm->delete_scheduler()->TEST_SetMaxTrashDBRatio(1.1);
 
   ASSERT_OK(TryReopen(options));
   // Create 4 files in L0
@@ -402,6 +403,7 @@ TEST_F(DBSSTTest, DeleteSchedulerMultipleDBPaths) {
       env_, nullptr, trash_dir, rate_bytes_per_sec, false, &s));
   ASSERT_OK(s);
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+  sfm->delete_scheduler()->TEST_SetMaxTrashDBRatio(1.1);
 
   DestroyAndReopen(options);
 
@@ -456,9 +458,14 @@ TEST_F(DBSSTTest, DestroyDBWithRateLimitedDelete) {
       [&](void* arg) { bg_delete_file++; });
   rocksdb::SyncPoint::GetInstance()->EnableProcessing();
 
+  Status s;
   Options options = CurrentOptions();
   options.disable_auto_compactions = true;
   options.env = env_;
+  std::string trash_dir = test::TmpDir(env_) + "/trash";
+  options.sst_file_manager.reset(
+      NewSstFileManager(env_, nullptr, trash_dir, 0, false, &s));
+  ASSERT_OK(s);
   DestroyAndReopen(options);
 
   // Create 4 files in L0
@@ -471,15 +478,12 @@ TEST_F(DBSSTTest, DestroyDBWithRateLimitedDelete) {
 
   // Close DB and destroy it using DeleteScheduler
   Close();
-  std::string trash_dir = test::TmpDir(env_) + "/trash";
-  int64_t rate_bytes_per_sec = 1024 * 1024;  // 1 Mb / Sec
-  Status s;
-  options.sst_file_manager.reset(NewSstFileManager(
-      env_, nullptr, trash_dir, rate_bytes_per_sec, false, &s));
-  ASSERT_OK(s);
-  ASSERT_OK(DestroyDB(dbname_, options));
 
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+
+  sfm->SetDeleteRateBytesPerSecond(1024 * 1024);
+  sfm->delete_scheduler()->TEST_SetMaxTrashDBRatio(1.1);
+  ASSERT_OK(DestroyDB(dbname_, options));
   sfm->WaitForEmptyTrash();
   // We have deleted the 4 sst files in the delete_scheduler
   ASSERT_EQ(bg_delete_file, 4);

util/delete_scheduler.cc

@@ -26,12 +26,14 @@ DeleteScheduler::DeleteScheduler(Env* env, const std::string& trash_dir,
                                  SstFileManagerImpl* sst_file_manager)
     : env_(env),
       trash_dir_(trash_dir),
+      total_trash_size_(0),
       rate_bytes_per_sec_(rate_bytes_per_sec),
       pending_files_(0),
       closing_(false),
       cv_(&mu_),
       info_log_(info_log),
       sst_file_manager_(sst_file_manager) {
+  assert(sst_file_manager != nullptr);
   bg_thread_.reset(
       new port::Thread(&DeleteScheduler::BackgroundEmptyTrash, this));
 }
@@ -49,11 +51,14 @@ DeleteScheduler::~DeleteScheduler() {
 
 Status DeleteScheduler::DeleteFile(const std::string& file_path) {
   Status s;
-  if (rate_bytes_per_sec_.load() <= 0) {
+  if (rate_bytes_per_sec_.load() <= 0 ||
-    // Rate limiting is disabled
+      total_trash_size_.load() >
+          sst_file_manager_->GetTotalSize() * max_trash_db_ratio_) {
+    // Rate limiting is disabled or trash size makes up more than
+    // max_trash_db_ratio_ (default 25%) of the total DB size
     TEST_SYNC_POINT("DeleteScheduler::DeleteFile");
     s = env_->DeleteFile(file_path);
-    if (s.ok() && sst_file_manager_) {
+    if (s.ok()) {
       sst_file_manager_->OnDeleteFile(file_path);
     }
     return s;
@@ -66,7 +71,7 @@ Status DeleteScheduler::DeleteFile(const std::string& file_path) {
     ROCKS_LOG_ERROR(info_log_, "Failed to move %s to trash directory (%s)",
                     file_path.c_str(), trash_dir_.c_str());
     s = env_->DeleteFile(file_path);
-    if (s.ok() && sst_file_manager_) {
+    if (s.ok()) {
       sst_file_manager_->OnDeleteFile(file_path);
     }
     return s;
@@ -123,8 +128,10 @@ Status DeleteScheduler::MoveToTrash(const std::string& file_path,
       break;
     }
   }
-  if (s.ok() && sst_file_manager_) {
+  if (s.ok()) {
-    sst_file_manager_->OnMoveFile(file_path, *path_in_trash);
+    uint64_t trash_file_size = 0;
+    sst_file_manager_->OnMoveFile(file_path, *path_in_trash, &trash_file_size);
+    total_trash_size_.fetch_add(trash_file_size);
   }
   return s;
 }
@@ -210,10 +217,9 @@ Status DeleteScheduler::DeleteTrashFile(const std::string& path_in_trash,
     *deleted_bytes = 0;
   } else {
     *deleted_bytes = file_size;
-    if (sst_file_manager_) {
+    total_trash_size_.fetch_sub(file_size);
     sst_file_manager_->OnDeleteFile(path_in_trash);
   }
-  }
 
   return s;
 }

util/delete_scheduler.h

@@ -59,6 +59,13 @@ class DeleteScheduler {
   // file_path => error status
   std::map<std::string, Status> GetBackgroundErrors();
 
+  uint64_t GetTotalTrashSize() { return total_trash_size_.load(); }
+
+  void TEST_SetMaxTrashDBRatio(double r) {
+    assert(r >= 0);
+    max_trash_db_ratio_ = r;
+  }
+
  private:
   Status MoveToTrash(const std::string& file_path, std::string* path_in_trash);
 
@@ -70,6 +77,8 @@ class DeleteScheduler {
   Env* env_;
   // Path to the trash directory
   std::string trash_dir_;
+  // total size of trash directory
+  std::atomic<uint64_t> total_trash_size_;
   // Maximum number of bytes that should be deleted per second
   std::atomic<int64_t> rate_bytes_per_sec_;
   // Mutex to protect queue_, pending_files_, bg_errors_, closing_
@@ -93,6 +102,9 @@ class DeleteScheduler {
   InstrumentedMutex file_move_mu_;
   Logger* info_log_;
   SstFileManagerImpl* sst_file_manager_;
+  // If the trash size constitutes for more than 25% of the total DB size
+  // we will start deleting new files passed to DeleteScheduler immediately
+  double max_trash_db_ratio_ = 0.25;
   static const uint64_t kMicrosInSecond = 1000 * 1000LL;
 };
 

util/delete_scheduler_test.cc

@@ -17,6 +17,7 @@
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
 #include "util/delete_scheduler.h"
+#include "util/sst_file_manager_impl.h"
 #include "util/string_util.h"
 #include "util/sync_point.h"
 #include "util/testharness.h"
@@ -61,20 +62,26 @@ class DeleteSchedulerTest : public testing::Test {
     std::string data(size, 'A');
     EXPECT_OK(f->Append(data));
     EXPECT_OK(f->Close());
+    sst_file_mgr_->OnAddFile(file_path);
     return file_path;
   }
 
   void NewDeleteScheduler() {
     ASSERT_OK(env_->CreateDirIfMissing(trash_dir_));
-    delete_scheduler_.reset(new DeleteScheduler(
+    sst_file_mgr_.reset(
-        env_, trash_dir_, rate_bytes_per_sec_, nullptr, nullptr));
+        new SstFileManagerImpl(env_, nullptr, trash_dir_, rate_bytes_per_sec_));
+    delete_scheduler_ = sst_file_mgr_->delete_scheduler();
+    // Tests in this file are for DeleteScheduler component and dont create any
+    // DBs, so we need to use set this value to 100% (instead of default 25%)
+    delete_scheduler_->TEST_SetMaxTrashDBRatio(1.1);
   }
 
   Env* env_;
   std::string dummy_files_dir_;
   std::string trash_dir_;
   int64_t rate_bytes_per_sec_;
-  std::shared_ptr<DeleteScheduler> delete_scheduler_;
+  DeleteScheduler* delete_scheduler_;
+  std::unique_ptr<SstFileManagerImpl> sst_file_mgr_;
 };
 
 // Test the basic functionality of DeleteScheduler (Rate Limiting).
@@ -389,7 +396,7 @@ TEST_F(DeleteSchedulerTest, DestructorWithNonEmptyQueue) {
 
   // Deleting 100 files will need >28 hours to delete
   // we will delete the DeleteScheduler while delete queue is not empty
-  delete_scheduler_.reset();
+  sst_file_mgr_.reset();
 
   ASSERT_LT(bg_delete_file, 100);
   ASSERT_GT(CountFilesInDir(trash_dir_), 0);
@@ -508,6 +515,42 @@ TEST_F(DeleteSchedulerTest, DISABLED_DynamicRateLimiting1) {
   }
 }
 
+TEST_F(DeleteSchedulerTest, ImmediateDeleteOn25PercDBSize) {
+  int bg_delete_file = 0;
+  int fg_delete_file = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DeleteScheduler::DeleteTrashFile:DeleteFile",
+      [&](void* arg) { bg_delete_file++; });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DeleteScheduler::DeleteFile", [&](void* arg) { fg_delete_file++; });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  int num_files = 100;  // 100 files
+  uint64_t file_size = 1024 * 10; // 100 KB as a file size
+  rate_bytes_per_sec_ = 1;  // 1 byte per sec (very slow trash delete)
+
+  NewDeleteScheduler();
+  delete_scheduler_->TEST_SetMaxTrashDBRatio(0.25);
+
+  std::vector<std::string> generated_files;
+  for (int i = 0; i < num_files; i++) {
+    std::string file_name = "file" + ToString(i) + ".data";
+    generated_files.push_back(NewDummyFile(file_name, file_size));
+  }
+
+  for (std::string& file_name : generated_files) {
+    delete_scheduler_->DeleteFile(file_name);
+  }
+
+  // When we end up with 24 files in trash we will start
+  // deleting new files immediately
+  ASSERT_EQ(fg_delete_file, 74);
+  ASSERT_EQ(CountFilesInDir(trash_dir_), 25);
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

util/sst_file_manager_impl.cc

@@ -51,9 +51,13 @@ Status SstFileManagerImpl::OnDeleteFile(const std::string& file_path) {
 }
 
 Status SstFileManagerImpl::OnMoveFile(const std::string& old_path,
-                                      const std::string& new_path) {
+                                      const std::string& new_path,
+                                      uint64_t* file_size) {
   {
     MutexLock l(&mu_);
+    if (file_size != nullptr) {
+      *file_size = tracked_files_[old_path];
+    }
     OnAddFileImpl(new_path, tracked_files_[old_path]);
     OnDeleteFileImpl(old_path);
   }

util/sst_file_manager_impl.h

@@ -39,7 +39,8 @@ class SstFileManagerImpl : public SstFileManager {
   Status OnDeleteFile(const std::string& file_path);
 
   // DB will call OnMoveFile whenever an sst file is move to a new path.
-  Status OnMoveFile(const std::string& old_path, const std::string& new_path);
+  Status OnMoveFile(const std::string& old_path, const std::string& new_path,
+                    uint64_t* file_size = nullptr);
 
   // Update the maximum allowed space that should be used by RocksDB, if
   // the total size of the SST files exceeds max_allowed_space, writes to
@@ -76,6 +77,8 @@ class SstFileManagerImpl : public SstFileManager {
   // destructor to be called.
   virtual void WaitForEmptyTrash();
 
+  DeleteScheduler* delete_scheduler() { return &delete_scheduler_; }
+
  private:
   // REQUIRES: mutex locked
   void OnAddFileImpl(const std::string& file_path, uint64_t file_size);