@nxtedition/rocksdb 7.1.20 → 7.1.21

package/deps/rocksdb/rocksdb/db/compaction/compaction_picker_fifo.cc

@@ -121,20 +121,45 @@ Compaction* FIFOCompactionPicker::PickTTLCompaction(
   return c;
 }
 
+// The size-based compaction picker for FIFO.
+//
+// When the entire column family size exceeds max_table_files_size, FIFO will
+// try to delete the oldest sst file(s) until the resulting column family size
+// is smaller than max_table_files_size.
+//
+// This function also takes care the case where a DB is migrating from level /
+// universal compaction to FIFO compaction.  During the migration, the column
+// family will also have non-L0 files while FIFO can only create L0 files.
+// In this case, this function will first purge the sst files in the bottom-
+// most non-empty level first, and the DB will eventually converge to the
+// regular FIFO case where there're only L0 files.  Note that during the
+// migration case, the purge order will only be an approximation of "FIFO"
+// as entries inside lower-level files might sometimes be newer than some
+// entries inside upper-level files.
 Compaction* FIFOCompactionPicker::PickSizeCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
     LogBuffer* log_buffer) {
-  const int kLevel0 = 0;
-  const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(kLevel0);
-  uint64_t total_size = GetTotalFilesSize(level_files);
+  // compute the total size and identify the last non-empty level
+  int last_level = 0;
+  uint64_t total_size = 0;
+  for (int level = 0; level < vstorage->num_levels(); ++level) {
+    auto level_size = GetTotalFilesSize(vstorage->LevelFiles(level));
+    total_size += level_size;
+    if (level_size > 0) {
+      last_level = level;
+    }
+  }
+  const std::vector<FileMetaData*>& last_level_files =
+      vstorage->LevelFiles(last_level);
 
-  if (total_size <=
-          mutable_cf_options.compaction_options_fifo.max_table_files_size ||
-      level_files.size() == 0) {
-    // total size not exceeded
+  if (last_level == 0 &&
+      total_size <=
+          mutable_cf_options.compaction_options_fifo.max_table_files_size) {
+    // total size not exceeded, try to find intra level 0 compaction if enabled
+    const std::vector<FileMetaData*>& level0_files = vstorage->LevelFiles(0);
     if (mutable_cf_options.compaction_options_fifo.allow_compaction &&
-        level_files.size() > 0) {
+        level0_files.size() > 0) {
       CompactionInputFiles comp_inputs;
       // try to prevent same files from being compacted multiple times, which
       // could produce large files that may never TTL-expire. Achieve this by
@@ -146,7 +171,7 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
               static_cast<uint64_t>(mutable_cf_options.write_buffer_size),
               1.1));
       if (FindIntraL0Compaction(
-              level_files,
+              level0_files,
               mutable_cf_options
                   .level0_file_num_compaction_trigger /* min_files_to_compact */
               ,
@@ -187,27 +212,58 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
 
   std::vector<CompactionInputFiles> inputs;
   inputs.emplace_back();
-  inputs[0].level = 0;
+  inputs[0].level = last_level;
 
-  for (auto ritr = level_files.rbegin(); ritr != level_files.rend(); ++ritr) {
-    auto f = *ritr;
-    total_size -= f->fd.file_size;
-    inputs[0].files.push_back(f);
-    char tmp_fsize[16];
-    AppendHumanBytes(f->fd.GetFileSize(), tmp_fsize, sizeof(tmp_fsize));
-    ROCKS_LOG_BUFFER(log_buffer,
-                     "[%s] FIFO compaction: picking file %" PRIu64
-                     " with size %s for deletion",
-                     cf_name.c_str(), f->fd.GetNumber(), tmp_fsize);
-    if (total_size <=
-        mutable_cf_options.compaction_options_fifo.max_table_files_size) {
-      break;
+  if (last_level == 0) {
+    // In L0, right-most files are the oldest files.
+    for (auto ritr = last_level_files.rbegin(); ritr != last_level_files.rend();
+         ++ritr) {
+      auto f = *ritr;
+      total_size -= f->fd.file_size;
+      inputs[0].files.push_back(f);
+      char tmp_fsize[16];
+      AppendHumanBytes(f->fd.GetFileSize(), tmp_fsize, sizeof(tmp_fsize));
+      ROCKS_LOG_BUFFER(log_buffer,
+                       "[%s] FIFO compaction: picking file %" PRIu64
+                       " with size %s for deletion",
+                       cf_name.c_str(), f->fd.GetNumber(), tmp_fsize);
+      if (total_size <=
+          mutable_cf_options.compaction_options_fifo.max_table_files_size) {
+        break;
+      }
+    }
+  } else {
+    // If the last level is non-L0, we actually don't know which file is
+    // logically the oldest since the file creation time only represents
+    // when this file was compacted to this level, which is independent
+    // to when the entries in this file were first inserted.
+    //
+    // As a result, we delete files from the left instead.  This means the sst
+    // file with the smallest key will be deleted first.  This design decision
+    // better serves a major type of FIFO use cases where smaller keys are
+    // associated with older data.
+    for (const auto& f : last_level_files) {
+      total_size -= f->fd.file_size;
+      inputs[0].files.push_back(f);
+      char tmp_fsize[16];
+      AppendHumanBytes(f->fd.GetFileSize(), tmp_fsize, sizeof(tmp_fsize));
+      ROCKS_LOG_BUFFER(log_buffer,
+                       "[%s] FIFO compaction: picking file %" PRIu64
+                       " with size %s for deletion",
+                       cf_name.c_str(), f->fd.GetNumber(), tmp_fsize);
+      if (total_size <=
+          mutable_cf_options.compaction_options_fifo.max_table_files_size) {
+        break;
+      }
     }
   }
 
   Compaction* c = new Compaction(
       vstorage, ioptions_, mutable_cf_options, mutable_db_options,
-      std::move(inputs), 0, 0, 0, 0, kNoCompression,
+      std::move(inputs), last_level,
+      /* target_file_size */ 0,
+      /* max_compaction_bytes */ 0,
+      /* output_path_id */ 0, kNoCompression,
       mutable_cf_options.compression_opts, Temperature::kUnknown,
       /* max_subcompactions */ 0, {}, /* is manual */ false,
       /* trim_ts */ "", vstorage->CompactionScore(0),
@@ -224,6 +280,13 @@ Compaction* FIFOCompactionPicker::PickCompactionToWarm(
     return nullptr;
   }
 
+  // PickCompactionToWarm is only triggered if there is no non-L0 files.
+  for (int level = 1; level < vstorage->num_levels(); ++level) {
+    if (GetTotalFilesSize(vstorage->LevelFiles(level)) > 0) {
+      return nullptr;
+    }
+  }
+
   const int kLevel0 = 0;
   const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(kLevel0);
 
@@ -327,8 +390,6 @@ Compaction* FIFOCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
     LogBuffer* log_buffer, SequenceNumber /*earliest_memtable_seqno*/) {
-  assert(vstorage->num_levels() == 1);
-
   Compaction* c = nullptr;
   if (mutable_cf_options.ttl > 0) {
     c = PickTTLCompaction(cf_name, mutable_cf_options, mutable_db_options,

package/deps/rocksdb/rocksdb/db/compaction/compaction_picker_level.cc

@@ -254,6 +254,25 @@ void LevelCompactionBuilder::SetupInitialFiles() {
   }
 
   // TTL Compaction
+  if (ioptions_.compaction_pri == kRoundRobin &&
+      !vstorage_->ExpiredTtlFiles().empty()) {
+    auto expired_files = vstorage_->ExpiredTtlFiles();
+    // the expired files list should already be sorted by level
+    start_level_ = expired_files.front().first;
+#ifndef NDEBUG
+    for (const auto& file : expired_files) {
+      assert(start_level_ <= file.first);
+    }
+#endif
+    if (start_level_ > 0) {
+      output_level_ = start_level_ + 1;
+      if (PickFileToCompact()) {
+        compaction_reason_ = CompactionReason::kRoundRobinTtl;
+        return;
+      }
+    }
+  }
+
   PickFileToCompact(vstorage_->ExpiredTtlFiles(), true);
   if (!start_level_inputs_.empty()) {
     compaction_reason_ = CompactionReason::kTtl;
@@ -318,7 +337,7 @@ void LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion() {
   // Constraint 3 (pre-calculate the ideal max bytes to compact)
   for (auto f : level_files) {
     if (!f->being_compacted) {
-      start_lvl_bytes_no_compacting += f->compensated_file_size;
+      start_lvl_bytes_no_compacting += f->fd.GetFileSize();
     }
   }
   if (start_lvl_bytes_no_compacting >
@@ -341,7 +360,7 @@ void LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion() {
     }
   }
   // Constraint 3
-  if (start_level_inputs_[0]->compensated_file_size >=
+  if (start_level_inputs_[0]->fd.GetFileSize() >=
       start_lvl_max_bytes_to_compact) {
     return;
   }
@@ -368,7 +387,7 @@ void LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion() {
 
     curr_bytes_to_compact = 0;
     for (auto start_lvl_f : tmp_start_level_inputs.files) {
-      curr_bytes_to_compact += start_lvl_f->compensated_file_size;
+      curr_bytes_to_compact += start_lvl_f->fd.GetFileSize();
     }
 
     // Check whether any output level files are locked
@@ -385,7 +404,7 @@ void LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion() {
 
     uint64_t start_lvl_curr_bytes_to_compact = curr_bytes_to_compact;
     for (auto output_lvl_f : output_level_inputs.files) {
-      curr_bytes_to_compact += output_lvl_f->compensated_file_size;
+      curr_bytes_to_compact += output_lvl_f->fd.GetFileSize();
     }
     if (curr_bytes_to_compact > mutable_cf_options_.max_compaction_bytes) {
       // Constraint 2

package/deps/rocksdb/rocksdb/db/compaction/compaction_picker_test.cc

@@ -1340,6 +1340,7 @@ TEST_F(CompactionPickerTest, CompactionPriMinOverlapping4) {
   ioptions_.compaction_pri = kMinOverlappingRatio;
   mutable_cf_options_.max_bytes_for_level_base = 10000000;
   mutable_cf_options_.max_bytes_for_level_multiplier = 10;
+  mutable_cf_options_.ignore_max_compaction_bytes_for_input = false;
 
   // file 7 and 8 over lap with the same file, but file 8 is smaller so
   // it will be picked.
@@ -2358,6 +2359,7 @@ TEST_F(CompactionPickerTest, IsBottommostLevelTest) {
 TEST_F(CompactionPickerTest, MaxCompactionBytesHit) {
   mutable_cf_options_.max_bytes_for_level_base = 1000000u;
   mutable_cf_options_.max_compaction_bytes = 800000u;
+  mutable_cf_options_.ignore_max_compaction_bytes_for_input = false;
   ioptions_.level_compaction_dynamic_level_bytes = false;
   NewVersionStorage(6, kCompactionStyleLevel);
   // A compaction should be triggered and pick file 2 and 5.
@@ -2384,6 +2386,7 @@ TEST_F(CompactionPickerTest, MaxCompactionBytesHit) {
 TEST_F(CompactionPickerTest, MaxCompactionBytesNotHit) {
   mutable_cf_options_.max_bytes_for_level_base = 800000u;
   mutable_cf_options_.max_compaction_bytes = 1000000u;
+  mutable_cf_options_.ignore_max_compaction_bytes_for_input = false;
   ioptions_.level_compaction_dynamic_level_bytes = false;
   NewVersionStorage(6, kCompactionStyleLevel);
   // A compaction should be triggered and pick file 2 and 5.
@@ -3656,6 +3659,63 @@ TEST_P(PerKeyPlacementCompactionPickerTest, NormalCompactionOverlapUniversal) {
                 input_files, 5));
 }
 
+TEST_P(PerKeyPlacementCompactionPickerTest, PenultimateOverlapUniversal) {
+  // This test is make sure the Tiered compaction would lock whole range of
+  // both output level and penultimate level
+  if (enable_per_key_placement_) {
+    ioptions_.preclude_last_level_data_seconds = 10000;
+  }
+
+  int num_levels = ioptions_.num_levels;
+  ioptions_.compaction_style = kCompactionStyleUniversal;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+  NewVersionStorage(num_levels, kCompactionStyleUniversal);
+
+  // L4:   [200, 220]  [230, 250]       [360, 380]
+  // L5:
+  // L6: [101,                    351]
+  Add(4, 40U, "200", "220", 60000000U);
+  Add(4, 41U, "230", "250", 60000000U);
+  Add(4, 42U, "360", "380", 60000000U);
+  Add(6, 50U, "101", "351", 60000000U);
+  UpdateVersionStorageInfo();
+
+  // the existing compaction is the 1st L4 file + L6 file
+  // then compaction of the 2nd L4 file to L5 (penultimate level) is overlapped
+  // when the tiered compaction feature is on.
+  CompactionOptions comp_options;
+  std::unordered_set<uint64_t> input_set;
+  input_set.insert(40);
+  input_set.insert(50);
+  std::vector<CompactionInputFiles> input_files;
+  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
+      &input_files, &input_set, vstorage_.get(), comp_options));
+
+  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
+      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+      mutable_db_options_, 0));
+
+  input_set.clear();
+  input_files.clear();
+  input_set.insert(41);
+  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
+      &input_files, &input_set, vstorage_.get(), comp_options));
+
+  ASSERT_EQ(enable_per_key_placement_,
+            universal_compaction_picker.FilesRangeOverlapWithCompaction(
+                input_files, 5));
+
+  // compacting the 3rd L4 file is always safe:
+  input_set.clear();
+  input_files.clear();
+  input_set.insert(42);
+  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
+      &input_files, &input_set, vstorage_.get(), comp_options));
+
+  ASSERT_FALSE(universal_compaction_picker.FilesRangeOverlapWithCompaction(
+      input_files, 5));
+}
+
 INSTANTIATE_TEST_CASE_P(PerKeyPlacementCompactionPickerTest,
                         PerKeyPlacementCompactionPickerTest, ::testing::Bool());
 
@@ -3664,6 +3724,7 @@ INSTANTIATE_TEST_CASE_P(PerKeyPlacementCompactionPickerTest,
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {
+  ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
 }

package/deps/rocksdb/rocksdb/db/compaction/tiered_compaction_test.cc

@@ -11,6 +11,8 @@
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/listener.h"
+#include "rocksdb/utilities/debug.h"
+#include "test_util/mock_time_env.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -416,36 +418,18 @@ TEST_P(TieredCompactionTest, RangeBasedTieredStorageUniversal) {
   ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
   ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
 
-  // Add 2 keys in higher level, but in separated files, the keys within that
-  // range should be moved up to the penultimate level
+  // Add 2 keys in higher level, but in separated files, all keys can be moved
+  // up if it's hot
   ASSERT_OK(Put(Key(0), "value" + std::to_string(0)));
   ASSERT_OK(Flush());
   ASSERT_OK(Put(Key(50), "value" + std::to_string(0)));
   ASSERT_OK(Flush());
-  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-  ASSERT_EQ("0,0,0,0,0,1,1", FilesPerLevel());
-
-  // Add an SST with a key range cover all the data we want to move from the
-  // last level to the penultimate level
-  ASSERT_OK(Put(Key(0), "value" + std::to_string(0)));
-  ASSERT_OK(Put(Key(99), "value" + std::to_string(0)));
-  ASSERT_OK(Flush());
-
-  ResetAllStats(expect_stats, expect_pl_stats);
-
   ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
   ASSERT_EQ("0,0,0,0,0,1", FilesPerLevel());
+
   ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
   ASSERT_EQ(GetSstSizeHelper(Temperature::kCold), 0);
 
-  last_stats.Add(kBasicCompStats);
-  last_stats.ResetCompactionReason(CompactionReason::kManualCompaction);
-  last_stats.bytes_read_output_level = kHasValue;
-  last_stats.num_input_files_in_output_level = kHasValue;
-  expect_pl_stats.Add(kBasicPerKeyPlacementCompStats);
-  expect_pl_stats.ResetCompactionReason(CompactionReason::kManualCompaction);
-  VerifyCompactionStats(expect_stats, expect_pl_stats);
-
   // change to only 1 key cold, to test compaction could stop even it matches
   // size amp compaction threshold
   {
@@ -1215,6 +1199,295 @@ TEST_P(TieredCompactionTest, RangeBasedTieredStorageLevel) {
 INSTANTIATE_TEST_CASE_P(TieredCompactionTest, TieredCompactionTest,
                         testing::Bool());
 
+class PrecludeLastLevelTest : public DBTestBase {
+ public:
+  PrecludeLastLevelTest()
+      : DBTestBase("preclude_last_level_test", /*env_do_fsync=*/false) {
+    mock_clock_ = std::make_shared<MockSystemClock>(env_->GetSystemClock());
+    mock_env_ = std::make_unique<CompositeEnvWrapper>(env_, mock_clock_);
+  }
+
+ protected:
+  std::unique_ptr<Env> mock_env_;
+  std::shared_ptr<MockSystemClock> mock_clock_;
+
+  void SetUp() override {
+    mock_clock_->InstallTimedWaitFixCallback();
+    SyncPoint::GetInstance()->SetCallBack(
+        "DBImpl::StartPeriodicTaskScheduler:Init", [&](void* arg) {
+          auto periodic_task_scheduler_ptr =
+              reinterpret_cast<PeriodicTaskScheduler*>(arg);
+          periodic_task_scheduler_ptr->TEST_OverrideTimer(mock_clock_.get());
+        });
+    mock_clock_->SetCurrentTime(0);
+  }
+};
+
+TEST_F(PrecludeLastLevelTest, MigrationFromPreserveTimeManualCompaction) {
+  const int kNumTrigger = 4;
+  const int kNumLevels = 7;
+  const int kNumKeys = 100;
+  const int kKeyPerSec = 10;
+
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.preserve_internal_time_seconds = 10000;
+  options.env = mock_env_.get();
+  options.level0_file_num_compaction_trigger = kNumTrigger;
+  options.num_levels = kNumLevels;
+  DestroyAndReopen(options);
+
+  // pass some time first, otherwise the first a few keys write time are going
+  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
+  dbfull()->TEST_WaitForPeridicTaskRun(
+      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
+
+  int sst_num = 0;
+  // Write files that are overlap and enough to trigger compaction
+  for (; sst_num < kNumTrigger; sst_num++) {
+    for (int i = 0; i < kNumKeys; i++) {
+      ASSERT_OK(Put(Key(sst_num * (kNumKeys - 1) + i), "value"));
+      dbfull()->TEST_WaitForPeridicTaskRun([&] {
+        mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec));
+      });
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_OK(dbfull()->WaitForCompact(true));
+
+  // all data is pushed to the last level
+  ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
+
+  // enable preclude feature
+  options.preclude_last_level_data_seconds = 10000;
+  options.last_level_temperature = Temperature::kCold;
+  Reopen(options);
+
+  // all data is hot, even they're in the last level
+  ASSERT_EQ(GetSstSizeHelper(Temperature::kCold), 0);
+  ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+
+  // Generate a sstable and trigger manual compaction
+  ASSERT_OK(Put(Key(10), "value"));
+  ASSERT_OK(Flush());
+
+  CompactRangeOptions cro;
+  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
+
+  // all data is moved up to the penultimate level
+  ASSERT_EQ("0,0,0,0,0,1", FilesPerLevel());
+  ASSERT_EQ(GetSstSizeHelper(Temperature::kCold), 0);
+  ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+
+  // close explicitly, because the env is local variable which will be released
+  // first.
+  Close();
+}
+
+TEST_F(PrecludeLastLevelTest, MigrationFromPreserveTimeAutoCompaction) {
+  const int kNumTrigger = 4;
+  const int kNumLevels = 7;
+  const int kNumKeys = 100;
+  const int kKeyPerSec = 10;
+
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.preserve_internal_time_seconds = 10000;
+  options.env = mock_env_.get();
+  options.level0_file_num_compaction_trigger = kNumTrigger;
+  options.num_levels = kNumLevels;
+  DestroyAndReopen(options);
+
+  // pass some time first, otherwise the first a few keys write time are going
+  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
+  dbfull()->TEST_WaitForPeridicTaskRun(
+      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
+
+  int sst_num = 0;
+  // Write files that are overlap and enough to trigger compaction
+  for (; sst_num < kNumTrigger; sst_num++) {
+    for (int i = 0; i < kNumKeys; i++) {
+      ASSERT_OK(Put(Key(sst_num * (kNumKeys - 1) + i), "value"));
+      dbfull()->TEST_WaitForPeridicTaskRun([&] {
+        mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec));
+      });
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_OK(dbfull()->WaitForCompact(true));
+
+  // all data is pushed to the last level
+  ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
+
+  // enable preclude feature
+  options.preclude_last_level_data_seconds = 10000;
+  options.last_level_temperature = Temperature::kCold;
+  // make sure it won't trigger Size Amp compaction, unlike normal Size Amp
+  // compaction which is typically a last level compaction, when tiered Storage
+  // ("preclude_last_level") is enabled, size amp won't include the last level.
+  // As the last level would be in cold tier and the size would not be a
+  // problem, which also avoid frequent hot to cold storage compaction.
+  options.compaction_options_universal.max_size_amplification_percent = 400;
+  Reopen(options);
+
+  // all data is hot, even they're in the last level
+  ASSERT_EQ(GetSstSizeHelper(Temperature::kCold), 0);
+  ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+
+  // Write more data, but still all hot until the 10th SST, as:
+  // write a key every 10 seconds, 100 keys per SST, each SST takes 1000 seconds
+  // The preclude_last_level_data_seconds is 10k
+  Random rnd(301);
+  for (; sst_num < kNumTrigger * 2 - 1; sst_num++) {
+    for (int i = 0; i < kNumKeys; i++) {
+      // the value needs to be big enough to trigger full compaction
+      ASSERT_OK(Put(Key(sst_num * (kNumKeys - 1) + i), rnd.RandomString(100)));
+      dbfull()->TEST_WaitForPeridicTaskRun([&] {
+        mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec));
+      });
+    }
+    ASSERT_OK(Flush());
+    ASSERT_OK(dbfull()->WaitForCompact(true));
+  }
+
+  // all data is moved up to the penultimate level
+  ASSERT_EQ("0,0,0,0,0,1", FilesPerLevel());
+  ASSERT_EQ(GetSstSizeHelper(Temperature::kCold), 0);
+  ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+
+  // close explicitly, because the env is local variable which will be released
+  // first.
+  Close();
+}
+
+TEST_F(PrecludeLastLevelTest, MigrationFromPreserveTimePartial) {
+  const int kNumTrigger = 4;
+  const int kNumLevels = 7;
+  const int kNumKeys = 100;
+  const int kKeyPerSec = 10;
+
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.preserve_internal_time_seconds = 2000;
+  options.env = mock_env_.get();
+  options.level0_file_num_compaction_trigger = kNumTrigger;
+  options.num_levels = kNumLevels;
+  DestroyAndReopen(options);
+
+  // pass some time first, otherwise the first a few keys write time are going
+  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
+  dbfull()->TEST_WaitForPeridicTaskRun(
+      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
+
+  int sst_num = 0;
+  // Write files that are overlap and enough to trigger compaction
+  for (; sst_num < kNumTrigger; sst_num++) {
+    for (int i = 0; i < kNumKeys; i++) {
+      ASSERT_OK(Put(Key(sst_num * (kNumKeys - 1) + i), "value"));
+      dbfull()->TEST_WaitForPeridicTaskRun([&] {
+        mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec));
+      });
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_OK(dbfull()->WaitForCompact(true));
+
+  // all data is pushed to the last level
+  ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
+
+  std::vector<KeyVersion> key_versions;
+  ASSERT_OK(GetAllKeyVersions(db_, Slice(), Slice(),
+                              std::numeric_limits<size_t>::max(),
+                              &key_versions));
+
+  // make sure there're more than 300 keys and first 100 keys are having seqno
+  // zeroed out, the last 100 key seqno not zeroed out
+  ASSERT_GT(key_versions.size(), 300);
+  for (int i = 0; i < 100; i++) {
+    ASSERT_EQ(key_versions[i].sequence, 0);
+  }
+  auto rit = key_versions.rbegin();
+  for (int i = 0; i < 100; i++) {
+    ASSERT_GT(rit->sequence, 0);
+    rit++;
+  }
+
+  // enable preclude feature
+  options.preclude_last_level_data_seconds = 2000;
+  options.last_level_temperature = Temperature::kCold;
+  Reopen(options);
+
+  // Generate a sstable and trigger manual compaction
+  ASSERT_OK(Put(Key(10), "value"));
+  ASSERT_OK(Flush());
+
+  CompactRangeOptions cro;
+  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
+
+  // some data are moved up, some are not
+  ASSERT_EQ("0,0,0,0,0,1,1", FilesPerLevel());
+  ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
+  ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+
+  Close();
+}
+
+TEST_F(PrecludeLastLevelTest, SmallPrecludeTime) {
+  const int kNumTrigger = 4;
+  const int kNumLevels = 7;
+  const int kNumKeys = 100;
+
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.preclude_last_level_data_seconds = 60;
+  options.preserve_internal_time_seconds = 0;
+  options.env = mock_env_.get();
+  options.level0_file_num_compaction_trigger = kNumTrigger;
+  options.num_levels = kNumLevels;
+  options.last_level_temperature = Temperature::kCold;
+  DestroyAndReopen(options);
+
+  Random rnd(301);
+
+  dbfull()->TEST_WaitForPeridicTaskRun([&] {
+    mock_clock_->MockSleepForSeconds(static_cast<int>(rnd.Uniform(10) + 1));
+  });
+
+  for (int i = 0; i < kNumKeys; i++) {
+    ASSERT_OK(Put(Key(i), rnd.RandomString(100)));
+    dbfull()->TEST_WaitForPeridicTaskRun([&] {
+      mock_clock_->MockSleepForSeconds(static_cast<int>(rnd.Uniform(2)));
+    });
+  }
+  ASSERT_OK(Flush());
+
+  TablePropertiesCollection tables_props;
+  ASSERT_OK(dbfull()->GetPropertiesOfAllTables(&tables_props));
+  ASSERT_EQ(tables_props.size(), 1);
+  ASSERT_FALSE(tables_props.begin()->second->seqno_to_time_mapping.empty());
+  SeqnoToTimeMapping tp_mapping;
+  ASSERT_OK(
+      tp_mapping.Add(tables_props.begin()->second->seqno_to_time_mapping));
+  ASSERT_OK(tp_mapping.Sort());
+  ASSERT_FALSE(tp_mapping.Empty());
+  auto seqs = tp_mapping.TEST_GetInternalMapping();
+  ASSERT_FALSE(seqs.empty());
+
+  // Wait more than preclude_last_level time, then make sure all the data is
+  // compacted to the last level even there's no write (no seqno -> time
+  // information was flushed to any SST).
+  mock_clock_->MockSleepForSeconds(100);
+
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
+  ASSERT_EQ(GetSstSizeHelper(Temperature::kUnknown), 0);
+  ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
+
+  Close();
+}
+
 #endif  // !defined(ROCKSDB_LITE)
 
 }  // namespace ROCKSDB_NAMESPACE

package/deps/rocksdb/rocksdb/db/comparator_db_test.cc

@@ -673,6 +673,7 @@ TEST_P(ComparatorDBTest, SeparatorSuccessorRandomizeTest) {
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {
+  ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
 }

package/deps/rocksdb/rocksdb/db/cuckoo_table_db_test.cc

@@ -345,6 +345,7 @@ TEST_F(CuckooTableDBTest, AdaptiveTable) {
 
 int main(int argc, char** argv) {
   if (ROCKSDB_NAMESPACE::port::kLittleEndian) {
+    ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
     ::testing::InitGoogleTest(&argc, argv);
     return RUN_ALL_TESTS();
   } else {