Add property `LiveSstFilesSizeAtTemperature` for tiered storage (#8644)

Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/8644

Reviewed By: siying, zhichao-cao

Differential Revision: D30236535

Pulled By: jay-zhuang

fbshipit-source-id: 1758d1c46d83a5087560fb63d53a016bf999da81

diff --git a/HISTORY.md b/HISTORY.md
index e68d6dc1f8b786997eff21b83173992c4d2fa655..8fe057683308954b02a3ac0c8274fedc262f46c4 100644 (file)
--- a/HISTORY.md
+++ b/HISTORY.md
@@ -16,6 +16,7 @@
 * Add CompactionOptionsFIFO.age_for_warm, which allows RocksDB to move old files to warm tier in FIFO compactions. Note that file temperature is still an experimental feature.
 * Add a comment to suggest btrfs user to disable file preallocation by setting `options.allow_fallocate=false`.
 * Fast forward option in Trace replay changed to double type to allow replaying at a lower speed, by settings the value between 0 and 1. This option can be set via `ReplayOptions` in `Replayer::Replay()`, or via `--trace_replay_fast_forward` in db_bench.
+* Add property `LiveSstFilesSizeAtTemperature` to retrieve sst file size at different temperature.
 
 ## Public API change
 * Added APIs to decode and replay trace file via Replayer class. Added `DB::NewDefaultReplayer()` to create a default Replayer instance. Added `TraceReader::Reset()` to restart reading a trace file. Created trace_record.h and utilities/replayer.h files to access decoded Trace records and replay them.

diff --git a/db/db_test2.cc b/db/db_test2.cc
index d55f4a449fcd1ab989a6d1cfea17e5affcb28f04..a7f82db1bac9319f320513edc6e8a1dfd5fd4d4b 100644 (file)
--- a/db/db_test2.cc
+++ b/db/db_test2.cc
@@ -27,6 +27,19 @@ namespace ROCKSDB_NAMESPACE {
 class DBTest2 : public DBTestBase {
  public:
   DBTest2() : DBTestBase("db_test2", /*env_do_fsync=*/true) {}
+
+ protected:
+#ifndef ROCKSDB_LITE
+  uint64_t GetSstSizeHelper(Temperature temperature) {
+    std::string prop;
+    bool s =
+        dbfull()->GetProperty(DB::Properties::kLiveSstFilesSizeAtTemperature +
+                                  ToString(static_cast<uint8_t>(temperature)),
+                              &prop);
+    assert(s);
+    return static_cast<uint64_t>(std::atoi(prop.c_str()));
+  }
+#endif  // ROCKSDB_LITE
 };
 
 #ifndef ROCKSDB_LITE
@@ -6127,6 +6140,13 @@ TEST_F(DBTest2, BottommostTemperature) {
   options.level0_file_num_compaction_trigger = 2;
   Reopen(options);
 
+  auto size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kHot);
+  ASSERT_EQ(size, 0);
+
   ASSERT_OK(Put("foo", "bar"));
   ASSERT_OK(Put("bar", "bar"));
   ASSERT_OK(Flush());
@@ -6139,6 +6159,10 @@ TEST_F(DBTest2, BottommostTemperature) {
   db_->GetColumnFamilyMetaData(&metadata);
   ASSERT_EQ(1, metadata.file_count);
   ASSERT_EQ(Temperature::kWarm, metadata.levels[1].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_GT(size, 0);
 
   // non-bottommost file still has unknown temperature
   ASSERT_OK(Put("foo", "bar"));
@@ -6147,6 +6171,10 @@ TEST_F(DBTest2, BottommostTemperature) {
   db_->GetColumnFamilyMetaData(&metadata);
   ASSERT_EQ(2, metadata.file_count);
   ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_GT(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_GT(size, 0);
 
   // reopen and check the information is persisted
   Reopen(options);
@@ -6154,6 +6182,21 @@ TEST_F(DBTest2, BottommostTemperature) {
   ASSERT_EQ(2, metadata.file_count);
   ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[0].temperature);
   ASSERT_EQ(Temperature::kWarm, metadata.levels[1].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_GT(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_GT(size, 0);
+
+  // check other non-exist temperatures
+  size = GetSstSizeHelper(Temperature::kHot);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kCold);
+  ASSERT_EQ(size, 0);
+  std::string prop;
+  ASSERT_TRUE(dbfull()->GetProperty(
+      DB::Properties::kLiveSstFilesSizeAtTemperature + std::to_string(22),
+      &prop));
+  ASSERT_EQ(std::atoi(prop.c_str()), 0);
 }
 
 TEST_F(DBTest2, BottommostTemperatureUniversal) {
@@ -6167,6 +6210,13 @@ TEST_F(DBTest2, BottommostTemperatureUniversal) {
 
   DestroyAndReopen(options);
 
+  auto size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kHot);
+  ASSERT_EQ(size, 0);
+
   for (int i = 0; i < kTriggerNum; i++) {
     ASSERT_OK(Put("foo", "bar"));
     ASSERT_OK(Put("bar", "bar"));
@@ -6179,6 +6229,10 @@ TEST_F(DBTest2, BottommostTemperatureUniversal) {
   ASSERT_EQ(1, metadata.file_count);
   ASSERT_EQ(Temperature::kUnknown,
             metadata.levels[kBottommostLevel].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_GT(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_EQ(size, 0);
 
   ASSERT_OK(Put("foo", "bar"));
   ASSERT_OK(Put("bar", "bar"));
@@ -6187,6 +6241,10 @@ TEST_F(DBTest2, BottommostTemperatureUniversal) {
   db_->GetColumnFamilyMetaData(&metadata);
   ASSERT_EQ(2, metadata.file_count);
   ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_GT(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_EQ(size, 0);
 
   // Update bottommost temperature
   options.bottommost_temperature = Temperature::kWarm;
@@ -6195,6 +6253,10 @@ TEST_F(DBTest2, BottommostTemperatureUniversal) {
   // Should not impact existing ones
   ASSERT_EQ(Temperature::kUnknown,
             metadata.levels[kBottommostLevel].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_GT(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_EQ(size, 0);
 
   // new generated file should have the new settings
   ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
@@ -6202,6 +6264,10 @@ TEST_F(DBTest2, BottommostTemperatureUniversal) {
   ASSERT_EQ(1, metadata.file_count);
   ASSERT_EQ(Temperature::kWarm,
             metadata.levels[kBottommostLevel].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_GT(size, 0);
 
   // non-bottommost file still has unknown temperature
   ASSERT_OK(Put("foo", "bar"));
@@ -6211,6 +6277,21 @@ TEST_F(DBTest2, BottommostTemperatureUniversal) {
   db_->GetColumnFamilyMetaData(&metadata);
   ASSERT_EQ(2, metadata.file_count);
   ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[0].temperature);
+  size = GetSstSizeHelper(Temperature::kUnknown);
+  ASSERT_GT(size, 0);
+  size = GetSstSizeHelper(Temperature::kWarm);
+  ASSERT_GT(size, 0);
+
+  // check other non-exist temperatures
+  size = GetSstSizeHelper(Temperature::kHot);
+  ASSERT_EQ(size, 0);
+  size = GetSstSizeHelper(Temperature::kCold);
+  ASSERT_EQ(size, 0);
+  std::string prop;
+  ASSERT_TRUE(dbfull()->GetProperty(
+      DB::Properties::kLiveSstFilesSizeAtTemperature + std::to_string(22),
+      &prop));
+  ASSERT_EQ(std::atoi(prop.c_str()), 0);
 }
 #endif  // ROCKSDB_LITE
 

diff --git a/db/internal_stats.cc b/db/internal_stats.cc
index 2fe2e644aae1c84ad560a95d9af95da7eb5b224e..a91883975dd3d7dbdff5ce01716b5eca3ceee07f 100644 (file)
--- a/db/internal_stats.cc
+++ b/db/internal_stats.cc
@@ -262,6 +262,8 @@ static const std::string min_obsolete_sst_number_to_keep_str =
 static const std::string base_level_str = "base-level";
 static const std::string total_sst_files_size = "total-sst-files-size";
 static const std::string live_sst_files_size = "live-sst-files-size";
+static const std::string live_sst_files_size_at_temperature =
+    "live-sst-files-size-at-temperature";
 static const std::string estimate_pending_comp_bytes =
     "estimate-pending-compaction-bytes";
 static const std::string aggregated_table_properties =
@@ -369,6 +371,8 @@ const std::string DB::Properties::kBlockCachePinnedUsage =
     rocksdb_prefix + block_cache_pinned_usage;
 const std::string DB::Properties::kOptionsStatistics =
     rocksdb_prefix + options_statistics;
+const std::string DB::Properties::kLiveSstFilesSizeAtTemperature =
+    rocksdb_prefix + live_sst_files_size_at_temperature;
 
 const std::unordered_map<std::string, DBPropertyInfo>
     InternalStats::ppt_name_to_info = {
@@ -482,6 +486,9 @@ const std::unordered_map<std::string, DBPropertyInfo>
         {DB::Properties::kLiveSstFilesSize,
          {false, nullptr, &InternalStats::HandleLiveSstFilesSize, nullptr,
           nullptr}},
+        {DB::Properties::kLiveSstFilesSizeAtTemperature,
+         {true, &InternalStats::HandleLiveSstFilesSizeAtTemperature, nullptr,
+          nullptr, nullptr}},
         {DB::Properties::kEstimatePendingCompactionBytes,
          {false, nullptr, &InternalStats::HandleEstimatePendingCompactionBytes,
           nullptr, nullptr}},
@@ -682,6 +689,28 @@ bool InternalStats::HandleBlockCacheEntryStatsMap(
   return true;
 }
 
+bool InternalStats::HandleLiveSstFilesSizeAtTemperature(std::string* value,
+                                                        Slice suffix) {
+  uint64_t temperature;
+  bool ok = ConsumeDecimalNumber(&suffix, &temperature) && suffix.empty();
+  if (!ok) {
+    return false;
+  }
+
+  uint64_t size = 0;
+  const auto* vstorage = cfd_->current()->storage_info();
+  for (int level = 0; level < vstorage->num_levels(); level++) {
+    for (const auto& file_meta : vstorage->LevelFiles(level)) {
+      if (static_cast<uint8_t>(file_meta->temperature) == temperature) {
+        size += file_meta->fd.GetFileSize();
+      }
+    }
+  }
+
+  *value = ToString(size);
+  return true;
+}
+
 const DBPropertyInfo* GetPropertyInfo(const Slice& property) {
   std::string ppt_name = GetPropertyNameAndArg(property).first.ToString();
   auto ppt_info_iter = InternalStats::ppt_name_to_info.find(ppt_name);

diff --git a/db/internal_stats.h b/db/internal_stats.h
index e3a869a63738d5dada256cb68a622cbf1b0ff134..b4cc6d8bd19c849238ab2a2bcc99a74e5249ff80 100644 (file)
--- a/db/internal_stats.h
+++ b/db/internal_stats.h
@@ -675,6 +675,7 @@ class InternalStats {
   bool HandleBlockCacheEntryStats(std::string* value, Slice suffix);
   bool HandleBlockCacheEntryStatsMap(std::map<std::string, std::string>* values,
                                      Slice suffix);
+  bool HandleLiveSstFilesSizeAtTemperature(std::string* value, Slice suffix);
   // Total number of background errors encountered. Every time a flush task
   // or compaction task fails, this counter is incremented. The failure can
   // be caused by any possible reason, including file system errors, out of

diff --git a/include/rocksdb/db.h b/include/rocksdb/db.h
index 1f715b3cfd9c4e121fd02b7de1636f964c5d63ff..264ff06f40378e7b0252dcd5272df552f0133cfd 100644 (file)
--- a/include/rocksdb/db.h
+++ b/include/rocksdb/db.h
@@ -914,6 +914,10 @@ class DB {
     //      files belong to the latest LSM tree.
     static const std::string kLiveSstFilesSize;
 
+    // "rocksdb.live_sst_files_size_at_temperature" - returns total size (bytes)
+    //      of SST files at all certain file temperature
+    static const std::string kLiveSstFilesSizeAtTemperature;
+
     //  "rocksdb.base-level" - returns number of level to which L0 data will be
     //      compacted.
     static const std::string kBaseLevel;