DeleteRange compaction statistics

Summary:
- "rocksdb.compaction.key.drop.range_del" - number of keys dropped during compaction due to a range tombstone covering them
- "rocksdb.compaction.range_del.drop.obsolete" - number of range tombstones dropped due to compaction to bottom level and no snapshot saving them
- s/CompactionIteratorStats/CompactionIterationStats/g since this class is no longer specific to CompactionIterator -- it's also updated for range tombstone iteration during compaction
- Move the above class into a separate .h file to avoid circular dependency.
Closes https://github.com/facebook/rocksdb/pull/1520

Differential Revision: D4187179

Pulled By: ajkr

fbshipit-source-id: 10c2103

db/compaction_iteration_stats.h

@@ -0,0 +1,29 @@
+//  Copyright (c) 2016-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#pragma once
+
+struct CompactionIterationStats {
+  // Compaction statistics
+  int64_t num_record_drop_user = 0;
+  int64_t num_record_drop_hidden = 0;
+  int64_t num_record_drop_obsolete = 0;
+  int64_t num_record_drop_range_del = 0;
+  int64_t num_range_del_drop_obsolete = 0;
+  uint64_t total_filter_time = 0;
+
+  // Input statistics
+  // TODO(noetzli): The stats are incomplete. They are lacking everything
+  // consumed by MergeHelper.
+  uint64_t num_input_records = 0;
+  uint64_t num_input_deletion_records = 0;
+  uint64_t num_input_corrupt_records = 0;
+  uint64_t total_input_raw_key_bytes = 0;
+  uint64_t total_input_raw_value_bytes = 0;
+
+  // Single-Delete diagnostics for exceptional situations
+  uint64_t num_single_del_fallthru = 0;
+  uint64_t num_single_del_mismatch = 0;
+};

db/compaction_iterator.cc

@@ -63,6 +63,8 @@ void CompactionIterator::ResetRecordCounts() {
   iter_stats_.num_record_drop_user = 0;
   iter_stats_.num_record_drop_hidden = 0;
   iter_stats_.num_record_drop_obsolete = 0;
+  iter_stats_.num_record_drop_range_del = 0;
+  iter_stats_.num_range_del_drop_obsolete = 0;
 }
 
 void CompactionIterator::SeekToFirst() {
@@ -424,6 +426,8 @@ void CompactionIterator::NextFromInput() {
       // 2. different snapshot stripe
       bool should_delete = range_del_agg_->ShouldDelete(key_);
       if (should_delete) {
+        ++iter_stats_.num_record_drop_hidden;
+        ++iter_stats_.num_record_drop_range_del;
         input_->Next();
       } else {
         valid_ = true;

db/compaction_iterator.h

@@ -12,6 +12,7 @@
 #include <vector>
 
 #include "db/compaction.h"
+#include "db/compaction_iteration_stats.h"
 #include "db/merge_helper.h"
 #include "db/pinned_iterators_manager.h"
 #include "db/range_del_aggregator.h"
@@ -20,27 +21,6 @@
 
 namespace rocksdb {
 
-struct CompactionIteratorStats {
-  // Compaction statistics
-  int64_t num_record_drop_user = 0;
-  int64_t num_record_drop_hidden = 0;
-  int64_t num_record_drop_obsolete = 0;
-  uint64_t total_filter_time = 0;
-
-  // Input statistics
-  // TODO(noetzli): The stats are incomplete. They are lacking everything
-  // consumed by MergeHelper.
-  uint64_t num_input_records = 0;
-  uint64_t num_input_deletion_records = 0;
-  uint64_t num_input_corrupt_records = 0;
-  uint64_t total_input_raw_key_bytes = 0;
-  uint64_t total_input_raw_value_bytes = 0;
-
-  // Single-Delete diagnostics for exceptional situations
-  uint64_t num_single_del_fallthru = 0;
-  uint64_t num_single_del_mismatch = 0;
-};
-
 class CompactionIterator {
  public:
   CompactionIterator(InternalIterator* input, const Comparator* cmp,
@@ -74,7 +54,7 @@ class CompactionIterator {
   const ParsedInternalKey& ikey() const { return ikey_; }
   bool Valid() const { return valid_; }
   const Slice& user_key() const { return current_user_key_; }
-  const CompactionIteratorStats& iter_stats() const { return iter_stats_; }
+  const CompactionIterationStats& iter_stats() const { return iter_stats_; }
 
  private:
   // Processes the input stream to find the next output
@@ -157,6 +137,6 @@ class CompactionIterator {
   // increasing so a later call to the function must be looking for a key that
   // is in or beyond the last file checked during the previous call
   std::vector<size_t> level_ptrs_;
-  CompactionIteratorStats iter_stats_;
+  CompactionIterationStats iter_stats_;
 };
 }  // namespace rocksdb

db/compaction_job.cc

@@ -768,8 +768,12 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
                sub_compact->ShouldStopBefore(
                    key, sub_compact->current_output_file_size) &&
                sub_compact->builder != nullptr) {
-      status = FinishCompactionOutputFile(input->status(), sub_compact,
-                                          range_del_agg.get());
+      CompactionIterationStats range_del_out_stats;
+      status =
+          FinishCompactionOutputFile(input->status(), sub_compact,
+                                     range_del_agg.get(), &range_del_out_stats);
+      RecordDroppedKeys(range_del_out_stats,
+                        &sub_compact->compaction_job_stats);
       if (!status.ok()) {
         break;
       }
@@ -861,8 +865,12 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       if (c_iter->Valid()) {
         next_key = &c_iter->key();
       }
+      CompactionIterationStats range_del_out_stats;
       status = FinishCompactionOutputFile(input_status, sub_compact,
-                                          range_del_agg.get(), next_key);
+                                          range_del_agg.get(),
+                                          &range_del_out_stats, next_key);
+      RecordDroppedKeys(range_del_out_stats,
+                        &sub_compact->compaction_job_stats);
       if (sub_compact->outputs.size() == 1) {
         // Use dictionary from first output file for compression of subsequent
         // files.
@@ -902,8 +910,11 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     status = OpenCompactionOutputFile(sub_compact);
   }
   if (status.ok() && sub_compact->builder != nullptr) {
-    status = FinishCompactionOutputFile(input->status(), sub_compact,
-                                        range_del_agg.get());
+    CompactionIterationStats range_del_out_stats;
+    status =
+        FinishCompactionOutputFile(input->status(), sub_compact,
+                                   range_del_agg.get(), &range_del_out_stats);
+    RecordDroppedKeys(range_del_out_stats, &sub_compact->compaction_job_stats);
   }
   if (status.ok()) {
     status = input->status();
@@ -929,7 +940,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 }
 
 void CompactionJob::RecordDroppedKeys(
-    const CompactionIteratorStats& c_iter_stats,
+    const CompactionIterationStats& c_iter_stats,
     CompactionJobStats* compaction_job_stats) {
   if (c_iter_stats.num_record_drop_user > 0) {
     RecordTick(stats_, COMPACTION_KEY_DROP_USER,
@@ -951,11 +962,20 @@ void CompactionJob::RecordDroppedKeys(
           c_iter_stats.num_record_drop_obsolete;
     }
   }
+  if (c_iter_stats.num_record_drop_range_del > 0) {
+    RecordTick(stats_, COMPACTION_KEY_DROP_RANGE_DEL,
+               c_iter_stats.num_record_drop_range_del);
+  }
+  if (c_iter_stats.num_range_del_drop_obsolete > 0) {
+    RecordTick(stats_, COMPACTION_RANGE_DEL_DROP_OBSOLETE,
+               c_iter_stats.num_range_del_drop_obsolete);
+  }
 }
 
 Status CompactionJob::FinishCompactionOutputFile(
     const Status& input_status, SubcompactionState* sub_compact,
     RangeDelAggregator* range_del_agg,
+    CompactionIterationStats* range_del_out_stats,
     const Slice* next_table_min_key /* = nullptr */) {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_COMPACTION_SYNC_FILE);
@@ -998,7 +1018,8 @@ Status CompactionJob::FinishCompactionOutputFile(
       upper_bound = sub_compact->end;
     }
     range_del_agg->AddToBuilder(sub_compact->builder.get(), lower_bound,
-                                upper_bound, meta, bottommost_level_);
+                                upper_bound, meta, range_del_out_stats,
+                                bottommost_level_);
   }
   const uint64_t current_entries = sub_compact->builder->NumEntries();
   meta->marked_for_compaction = sub_compact->builder->NeedCompact();

db/compaction_job.h

@@ -96,17 +96,18 @@ class CompactionJob {
   // kv-pairs
   void ProcessKeyValueCompaction(SubcompactionState* sub_compact);
 
-  Status FinishCompactionOutputFile(const Status& input_status,
-                                    SubcompactionState* sub_compact,
-                                    RangeDelAggregator* range_del_agg = nullptr,
-                                    const Slice* next_table_min_key = nullptr);
+  Status FinishCompactionOutputFile(
+      const Status& input_status, SubcompactionState* sub_compact,
+      RangeDelAggregator* range_del_agg,
+      CompactionIterationStats* range_del_out_stats,
+      const Slice* next_table_min_key = nullptr);
   Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options);
   void RecordCompactionIOStats();
   Status OpenCompactionOutputFile(SubcompactionState* sub_compact);
   void CleanupCompaction();
   void UpdateCompactionJobStats(
     const InternalStats::CompactionStats& stats) const;
-  void RecordDroppedKeys(const CompactionIteratorStats& c_iter_stats,
+  void RecordDroppedKeys(const CompactionIterationStats& c_iter_stats,
                          CompactionJobStats* compaction_job_stats = nullptr);
 
   void UpdateCompactionStats();

db/range_del_aggregator.cc

@@ -140,6 +140,7 @@ RangeDelAggregator::TombstoneMap& RangeDelAggregator::GetTombstoneMap(
 void RangeDelAggregator::AddToBuilder(
     TableBuilder* builder, const Slice* lower_bound, const Slice* upper_bound,
     FileMetaData* meta,
+    CompactionIterationStats* range_del_out_stats /* = nullptr */,
     bool bottommost_level /* = false */) {
   if (rep_ == nullptr) {
     return;
@@ -147,6 +148,10 @@ void RangeDelAggregator::AddToBuilder(
   auto stripe_map_iter = rep_->stripe_map_.begin();
   assert(stripe_map_iter != rep_->stripe_map_.end());
   if (bottommost_level) {
+    range_del_out_stats->num_range_del_drop_obsolete +=
+        static_cast<int64_t>(stripe_map_iter->second.size());
+    range_del_out_stats->num_record_drop_obsolete +=
+        static_cast<int64_t>(stripe_map_iter->second.size());
     // For the bottommost level, keys covered by tombstones in the first
     // (oldest) stripe have been compacted away, so the tombstones are obsolete.
     ++stripe_map_iter;

db/range_del_aggregator.h

@@ -9,6 +9,7 @@
 #include <string>
 #include <vector>
 
+#include "db/compaction_iteration_stats.h"
 #include "db/dbformat.h"
 #include "db/pinned_iterators_manager.h"
 #include "db/version_edit.h"
@@ -81,6 +82,7 @@ class RangeDelAggregator {
   //    compaction.
   void AddToBuilder(TableBuilder* builder, const Slice* lower_bound,
                     const Slice* upper_bound, FileMetaData* meta,
+                    CompactionIterationStats* range_del_out_stats = nullptr,
                     bool bottommost_level = false);
   bool IsEmpty();
 

include/rocksdb/statistics.h

@@ -95,12 +95,16 @@ enum Tickers : uint32_t {
 
   /**
    * COMPACTION_KEY_DROP_* count the reasons for key drop during compaction
-   * There are 3 reasons currently.
+   * There are 4 reasons currently.
    */
   COMPACTION_KEY_DROP_NEWER_ENTRY,  // key was written with a newer value.
+                                    // Also includes keys dropped for range del.
   COMPACTION_KEY_DROP_OBSOLETE,     // The key is obsolete.
+  COMPACTION_KEY_DROP_RANGE_DEL,    // key was covered by a range tombstone.
   COMPACTION_KEY_DROP_USER,  // user compaction function has dropped the key.
 
+  COMPACTION_RANGE_DEL_DROP_OBSOLETE,  // all keys in range were deleted.
+
   // Number of keys written to the database via the Put and Write call's
   NUMBER_KEYS_WRITTEN,
   // Number of Keys read,
@@ -251,7 +255,10 @@ const std::vector<std::pair<Tickers, std::string>> TickersNameMap = {
     {GET_HIT_L2_AND_UP, "rocksdb.l2andup.hit"},
     {COMPACTION_KEY_DROP_NEWER_ENTRY, "rocksdb.compaction.key.drop.new"},
     {COMPACTION_KEY_DROP_OBSOLETE, "rocksdb.compaction.key.drop.obsolete"},
+    {COMPACTION_KEY_DROP_RANGE_DEL, "rocksdb.compaction.key.drop.range_del"},
     {COMPACTION_KEY_DROP_USER, "rocksdb.compaction.key.drop.user"},
+    {COMPACTION_RANGE_DEL_DROP_OBSOLETE,
+     "rocksdb.compaction.range_del.drop.obsolete"},
     {NUMBER_KEYS_WRITTEN, "rocksdb.number.keys.written"},
     {NUMBER_KEYS_READ, "rocksdb.number.keys.read"},
     {NUMBER_KEYS_UPDATED, "rocksdb.number.keys.updated"},