@nxtedition/rocksdb 7.0.12 → 7.0.13

package/deps/rocksdb/rocksdb/db/c.cc

@@ -23,6 +23,7 @@
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
+#include "rocksdb/experimental.h"
 #include "rocksdb/filter_policy.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/memtablerep.h"
@@ -59,6 +60,7 @@ using ROCKSDB_NAMESPACE::Cache;
 using ROCKSDB_NAMESPACE::Checkpoint;
 using ROCKSDB_NAMESPACE::ColumnFamilyDescriptor;
 using ROCKSDB_NAMESPACE::ColumnFamilyHandle;
+using ROCKSDB_NAMESPACE::ColumnFamilyMetaData;
 using ROCKSDB_NAMESPACE::ColumnFamilyOptions;
 using ROCKSDB_NAMESPACE::CompactionFilter;
 using ROCKSDB_NAMESPACE::CompactionFilterFactory;
@@ -78,6 +80,7 @@ using ROCKSDB_NAMESPACE::FlushOptions;
 using ROCKSDB_NAMESPACE::InfoLogLevel;
 using ROCKSDB_NAMESPACE::IngestExternalFileOptions;
 using ROCKSDB_NAMESPACE::Iterator;
+using ROCKSDB_NAMESPACE::LevelMetaData;
 using ROCKSDB_NAMESPACE::LiveFileMetaData;
 using ROCKSDB_NAMESPACE::Logger;
 using ROCKSDB_NAMESPACE::LRUCacheOptions;
@@ -105,6 +108,7 @@ using ROCKSDB_NAMESPACE::Slice;
 using ROCKSDB_NAMESPACE::SliceParts;
 using ROCKSDB_NAMESPACE::SliceTransform;
 using ROCKSDB_NAMESPACE::Snapshot;
+using ROCKSDB_NAMESPACE::SstFileMetaData;
 using ROCKSDB_NAMESPACE::SstFileWriter;
 using ROCKSDB_NAMESPACE::Status;
 using ROCKSDB_NAMESPACE::TablePropertiesCollectorFactory;
@@ -178,6 +182,15 @@ struct rocksdb_cache_t {
 };
 struct rocksdb_livefiles_t       { std::vector<LiveFileMetaData> rep; };
 struct rocksdb_column_family_handle_t  { ColumnFamilyHandle* rep; };
+struct rocksdb_column_family_metadata_t {
+  ColumnFamilyMetaData rep;
+};
+struct rocksdb_level_metadata_t {
+  const LevelMetaData* rep;
+};
+struct rocksdb_sst_file_metadata_t {
+  const SstFileMetaData* rep;
+};
 struct rocksdb_envoptions_t      { EnvOptions        rep; };
 struct rocksdb_ingestexternalfileoptions_t  { IngestExternalFileOptions rep; };
 struct rocksdb_sstfilewriter_t   { SstFileWriter*    rep; };
@@ -1737,6 +1750,29 @@ void rocksdb_compact_range_cf(
       (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
 }
 
+void rocksdb_suggest_compact_range(rocksdb_t* db, const char* start_key,
+                                   size_t start_key_len, const char* limit_key,
+                                   size_t limit_key_len, char** errptr) {
+  Slice a, b;
+  Status s = ROCKSDB_NAMESPACE::experimental::SuggestCompactRange(
+      db->rep,
+      (start_key ? (a = Slice(start_key, start_key_len), &a) : nullptr),
+      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
+  SaveError(errptr, s);
+}
+
+void rocksdb_suggest_compact_range_cf(
+    rocksdb_t* db, rocksdb_column_family_handle_t* column_family,
+    const char* start_key, size_t start_key_len, const char* limit_key,
+    size_t limit_key_len, char** errptr) {
+  Slice a, b;
+  Status s = db->rep->SuggestCompactRange(
+      column_family->rep,
+      (start_key ? (a = Slice(start_key, start_key_len), &a) : nullptr),
+      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
+  SaveError(errptr, s);
+}
+
 void rocksdb_compact_range_opt(rocksdb_t* db, rocksdb_compactoptions_t* opt,
                                const char* start_key, size_t start_key_len,
                                const char* limit_key, size_t limit_key_len) {
@@ -3338,11 +3374,6 @@ unsigned char rocksdb_options_get_advise_random_on_open(
   return opt->rep.advise_random_on_open;
 }
 
-void rocksdb_options_set_experimental_mempurge_threshold(rocksdb_options_t* opt,
-                                                         double v) {
-  opt->rep.experimental_mempurge_threshold = v;
-}
-
 void rocksdb_options_set_access_hint_on_compaction_start(
     rocksdb_options_t* opt, int v) {
   switch(v) {
@@ -3528,6 +3559,16 @@ int rocksdb_options_get_max_background_flushes(rocksdb_options_t* opt) {
   return opt->rep.max_background_flushes;
 }
 
+void rocksdb_options_set_experimental_mempurge_threshold(rocksdb_options_t* opt,
+                                                         double v) {
+  opt->rep.experimental_mempurge_threshold = v;
+}
+
+double rocksdb_options_get_experimental_mempurge_threshold(
+    rocksdb_options_t* opt) {
+  return opt->rep.experimental_mempurge_threshold;
+}
+
 void rocksdb_options_set_max_log_file_size(rocksdb_options_t* opt, size_t v) {
   opt->rep.max_log_file_size = v;
 }
@@ -5131,6 +5172,119 @@ void rocksdb_delete_file_in_range_cf(
           (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr)));
 }
 
+/* MetaData */
+
+rocksdb_column_family_metadata_t* rocksdb_get_column_family_metadata(
+    rocksdb_t* db) {
+  rocksdb_column_family_metadata_t* meta = new rocksdb_column_family_metadata_t;
+  db->rep->GetColumnFamilyMetaData(&meta->rep);
+  return meta;
+}
+
+rocksdb_column_family_metadata_t* rocksdb_get_column_family_metadata_cf(
+    rocksdb_t* db, rocksdb_column_family_handle_t* column_family) {
+  rocksdb_column_family_metadata_t* meta = new rocksdb_column_family_metadata_t;
+  db->rep->GetColumnFamilyMetaData(column_family->rep, &meta->rep);
+  return meta;
+}
+
+void rocksdb_column_family_metadata_destroy(
+    rocksdb_column_family_metadata_t* cf_meta) {
+  delete cf_meta;
+}
+
+uint64_t rocksdb_column_family_metadata_get_size(
+    rocksdb_column_family_metadata_t* cf_meta) {
+  return cf_meta->rep.size;
+}
+
+size_t rocksdb_column_family_metadata_get_file_count(
+    rocksdb_column_family_metadata_t* cf_meta) {
+  return cf_meta->rep.file_count;
+}
+
+char* rocksdb_column_family_metadata_get_name(
+    rocksdb_column_family_metadata_t* cf_meta) {
+  return strdup(cf_meta->rep.name.c_str());
+}
+
+size_t rocksdb_column_family_metadata_get_level_count(
+    rocksdb_column_family_metadata_t* cf_meta) {
+  return cf_meta->rep.levels.size();
+}
+
+rocksdb_level_metadata_t* rocksdb_column_family_metadata_get_level_metadata(
+    rocksdb_column_family_metadata_t* cf_meta, size_t i) {
+  if (i >= cf_meta->rep.levels.size()) {
+    return NULL;
+  }
+  rocksdb_level_metadata_t* level_meta =
+      (rocksdb_level_metadata_t*)malloc(sizeof(rocksdb_level_metadata_t));
+  level_meta->rep = &cf_meta->rep.levels[i];
+
+  return level_meta;
+}
+
+void rocksdb_level_metadata_destroy(rocksdb_level_metadata_t* level_meta) {
+  // Only free the base pointer as its parent rocksdb_column_family_metadata_t
+  // has the ownership of its rep.
+  free(level_meta);
+}
+
+int rocksdb_level_metadata_get_level(rocksdb_level_metadata_t* level_meta) {
+  return level_meta->rep->level;
+}
+
+uint64_t rocksdb_level_metadata_get_size(rocksdb_level_metadata_t* level_meta) {
+  return level_meta->rep->size;
+}
+
+size_t rocksdb_level_metadata_get_file_count(
+    rocksdb_level_metadata_t* level_meta) {
+  return level_meta->rep->files.size();
+}
+
+rocksdb_sst_file_metadata_t* rocksdb_level_metadata_get_sst_file_metadata(
+    rocksdb_level_metadata_t* level_meta, size_t i) {
+  if (i >= level_meta->rep->files.size()) {
+    return nullptr;
+  }
+  rocksdb_sst_file_metadata_t* file_meta =
+      (rocksdb_sst_file_metadata_t*)malloc(sizeof(rocksdb_sst_file_metadata_t));
+  file_meta->rep = &level_meta->rep->files[i];
+  return file_meta;
+}
+
+void rocksdb_sst_file_metadata_destroy(rocksdb_sst_file_metadata_t* file_meta) {
+  // Only free the base pointer as its parent rocksdb_level_metadata_t
+  // has the ownership of its rep.
+  free(file_meta);
+}
+
+char* rocksdb_sst_file_metadata_get_relative_filename(
+    rocksdb_sst_file_metadata_t* file_meta) {
+  return strdup(file_meta->rep->relative_filename.c_str());
+}
+
+uint64_t rocksdb_sst_file_metadata_get_size(
+    rocksdb_sst_file_metadata_t* file_meta) {
+  return file_meta->rep->size;
+}
+
+char* rocksdb_sst_file_metadata_get_smallestkey(
+    rocksdb_sst_file_metadata_t* file_meta, size_t* key_len) {
+  *key_len = file_meta->rep->smallestkey.size();
+  return CopyString(file_meta->rep->smallestkey);
+}
+
+char* rocksdb_sst_file_metadata_get_largestkey(
+    rocksdb_sst_file_metadata_t* file_meta, size_t* key_len) {
+  *key_len = file_meta->rep->largestkey.size();
+  return CopyString(file_meta->rep->largestkey);
+}
+
+/* Transactions */
+
 rocksdb_transactiondb_options_t* rocksdb_transactiondb_options_create() {
   return new rocksdb_transactiondb_options_t;
 }

package/deps/rocksdb/rocksdb/db/c_test.c

@@ -272,6 +272,92 @@ static rocksdb_compactionfilter_t* CFilterCreate(
                                          CFilterName);
 }
 
+void CheckMetaData(rocksdb_column_family_metadata_t* cf_meta,
+                   const char* expected_cf_name) {
+  char* cf_name = rocksdb_column_family_metadata_get_name(cf_meta);
+  assert(strcmp(cf_name, expected_cf_name) == 0);
+  rocksdb_free(cf_name);
+
+  size_t cf_size = rocksdb_column_family_metadata_get_size(cf_meta);
+  assert(cf_size > 0);
+  size_t cf_file_count = rocksdb_column_family_metadata_get_size(cf_meta);
+  assert(cf_file_count > 0);
+
+  uint64_t total_level_size = 0;
+  size_t total_file_count = 0;
+  size_t level_count = rocksdb_column_family_metadata_get_level_count(cf_meta);
+  assert(level_count > 0);
+  for (size_t l = 0; l < level_count; ++l) {
+    rocksdb_level_metadata_t* level_meta =
+        rocksdb_column_family_metadata_get_level_metadata(cf_meta, l);
+    assert(level_meta);
+    assert(rocksdb_level_metadata_get_level(level_meta) >= (int)l);
+    uint64_t level_size = rocksdb_level_metadata_get_size(level_meta);
+    uint64_t file_size_in_level = 0;
+
+    size_t file_count = rocksdb_level_metadata_get_file_count(level_meta);
+    total_file_count += file_count;
+    for (size_t f = 0; f < file_count; ++f) {
+      rocksdb_sst_file_metadata_t* file_meta =
+          rocksdb_level_metadata_get_sst_file_metadata(level_meta, f);
+      assert(file_meta);
+
+      uint64_t file_size = rocksdb_sst_file_metadata_get_size(file_meta);
+      assert(file_size > 0);
+      file_size_in_level += file_size;
+
+      char* file_name =
+          rocksdb_sst_file_metadata_get_relative_filename(file_meta);
+      assert(file_name);
+      assert(strlen(file_name) > 0);
+      rocksdb_free(file_name);
+
+      size_t smallest_key_len;
+      char* smallest_key = rocksdb_sst_file_metadata_get_smallestkey(
+          file_meta, &smallest_key_len);
+      assert(smallest_key);
+      assert(smallest_key_len > 0);
+      size_t largest_key_len;
+      char* largest_key =
+          rocksdb_sst_file_metadata_get_largestkey(file_meta, &largest_key_len);
+      assert(largest_key);
+      assert(largest_key_len > 0);
+      rocksdb_free(smallest_key);
+      rocksdb_free(largest_key);
+
+      rocksdb_sst_file_metadata_destroy(file_meta);
+    }
+    assert(level_size == file_size_in_level);
+    total_level_size += level_size;
+    rocksdb_level_metadata_destroy(level_meta);
+  }
+  assert(total_file_count > 0);
+  assert(cf_size == total_level_size);
+}
+
+void GetAndCheckMetaData(rocksdb_t* db) {
+  rocksdb_column_family_metadata_t* cf_meta =
+      rocksdb_get_column_family_metadata(db);
+
+  CheckMetaData(cf_meta, "default");
+
+  rocksdb_column_family_metadata_destroy(cf_meta);
+}
+
+void GetAndCheckMetaDataCf(rocksdb_t* db,
+                           rocksdb_column_family_handle_t* handle,
+                           const char* cf_name) {
+  // Compact to make sure we have at least one sst file to obtain datadata.
+  rocksdb_compact_range_cf(db, handle, NULL, 0, NULL, 0);
+
+  rocksdb_column_family_metadata_t* cf_meta =
+      rocksdb_get_column_family_metadata_cf(db, handle);
+
+  CheckMetaData(cf_meta, cf_name);
+
+  rocksdb_column_family_metadata_destroy(cf_meta);
+}
+
 static rocksdb_t* CheckCompaction(rocksdb_t* db, rocksdb_options_t* options,
                                   rocksdb_readoptions_t* roptions,
                                   rocksdb_writeoptions_t* woptions) {
@@ -304,6 +390,11 @@ static rocksdb_t* CheckCompaction(rocksdb_t* db, rocksdb_options_t* options,
   CheckGet(db, roptions, "foo", "foovalue");
   CheckGet(db, roptions, "bar", NULL);
   CheckGet(db, roptions, "baz", "newbazvalue");
+
+  rocksdb_suggest_compact_range(db, "bar", 3, "foo", 3, &err);
+  GetAndCheckMetaData(db);
+  CheckNoError(err);
+
   return db;
 }
 
@@ -1296,6 +1387,9 @@ int main(int argc, char** argv) {
     CheckNoError(err);
     rocksdb_put_cf(db, woptions, handles[1], "foobar4", 7, "hello4", 6, &err);
     CheckNoError(err);
+    rocksdb_suggest_compact_range_cf(db, handles[1], "foo", 3, "foobar9", 7,
+                                     &err);
+    CheckNoError(err);
 
     rocksdb_flushoptions_t *flush_options = rocksdb_flushoptions_create();
     rocksdb_flushoptions_set_wait(flush_options, 1);
@@ -1443,6 +1537,8 @@ int main(int argc, char** argv) {
     CheckNoError(err);
     rocksdb_iter_destroy(iter);
 
+    GetAndCheckMetaDataCf(db, handles[1], cf_names[1]);
+
     rocksdb_drop_column_family(db, handles[1], &err);
     CheckNoError(err);
     for (i = 0; i < 2; i++) {
@@ -1831,6 +1927,10 @@ int main(int argc, char** argv) {
     rocksdb_options_set_wal_compression(o, 1);
     CheckCondition(1 == rocksdb_options_get_wal_compression(o));
 
+    rocksdb_options_set_experimental_mempurge_threshold(o, 29.0);
+    CheckCondition(29.0 ==
+                   rocksdb_options_get_experimental_mempurge_threshold(o));
+
     /* Blob Options */
     rocksdb_options_set_enable_blob_files(o, 1);
     CheckCondition(1 == rocksdb_options_get_enable_blob_files(o));
@@ -1961,6 +2061,8 @@ int main(int argc, char** argv) {
     CheckCondition(4 == rocksdb_options_get_bottommost_compression(copy));
     CheckCondition(2 == rocksdb_options_get_compaction_style(copy));
     CheckCondition(1 == rocksdb_options_get_atomic_flush(copy));
+    CheckCondition(29.0 ==
+                   rocksdb_options_get_experimental_mempurge_threshold(copy));
 
     // Copies should be independent.
     rocksdb_options_set_allow_ingest_behind(copy, 0);
@@ -2309,6 +2411,12 @@ int main(int argc, char** argv) {
     CheckCondition(0 == rocksdb_options_get_atomic_flush(copy));
     CheckCondition(1 == rocksdb_options_get_atomic_flush(o));
 
+    rocksdb_options_set_experimental_mempurge_threshold(copy, 229.0);
+    CheckCondition(229.0 ==
+                   rocksdb_options_get_experimental_mempurge_threshold(copy));
+    CheckCondition(29.0 ==
+                   rocksdb_options_get_experimental_mempurge_threshold(o));
+
     rocksdb_options_destroy(copy);
     rocksdb_options_destroy(o);
   }

package/deps/rocksdb/rocksdb/db/column_family.cc

@@ -550,7 +550,8 @@ ColumnFamilyData::ColumnFamilyData(
       prev_compaction_needed_bytes_(0),
       allow_2pc_(db_options.allow_2pc),
       last_memtable_id_(0),
-      db_paths_registered_(false) {
+      db_paths_registered_(false),
+      mempurge_used_(false) {
   if (id_ != kDummyColumnFamilyDataId) {
     // TODO(cc): RegisterDbPaths can be expensive, considering moving it
     // outside of this constructor which might be called with db mutex held.

package/deps/rocksdb/rocksdb/db/column_family.h

@@ -480,11 +480,8 @@ class ColumnFamilyData {
       const MutableCFOptions& mutable_cf_options,
       const ImmutableCFOptions& immutable_cf_options);
 
-  // Recalculate some small conditions, which are changed only during
-  // compaction, adding new memtable and/or
-  // recalculation of compaction score. These values are used in
-  // DBImpl::MakeRoomForWrite function to decide, if it need to make
-  // a write stall
+  // Recalculate some stall conditions, which are changed only during
+  // compaction, adding new memtable and/or recalculation of compaction score.
   WriteStallCondition RecalculateWriteStallConditions(
       const MutableCFOptions& mutable_cf_options);
 
@@ -529,6 +526,10 @@ class ColumnFamilyData {
 
   static const uint32_t kDummyColumnFamilyDataId;
 
+  // Keep track of whether the mempurge feature was ever used.
+  void SetMempurgeUsed() { mempurge_used_ = true; }
+  bool GetMempurgeUsed() { return mempurge_used_; }
+
  private:
   friend class ColumnFamilySet;
   ColumnFamilyData(uint32_t id, const std::string& name,
@@ -629,6 +630,7 @@ class ColumnFamilyData {
   // For charging memory usage of file metadata created for newly added files to
   // a Version associated with this CFD
   std::shared_ptr<CacheReservationManager> file_metadata_cache_res_mgr_;
+  bool mempurge_used_;
 };
 
 // ColumnFamilySet has interesting thread-safety requirements

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

@@ -279,6 +279,28 @@ Compaction::Compaction(
   }
 
   GetBoundaryKeys(vstorage, inputs_, &smallest_user_key_, &largest_user_key_);
+
+  // Every compaction regardless of any compaction reason may respect the
+  // existing compact cursor in the output level to split output files
+  InternalKey temp_split_key = InternalKey();
+  if (immutable_options_.compaction_style == kCompactionStyleLevel &&
+      immutable_options_.compaction_pri == kRoundRobin) {
+    const InternalKey cursor =
+        input_vstorage_->GetCompactCursors()[output_level_];
+    if (cursor.Valid()) {
+      const Slice& cursor_user_key = ExtractUserKey(cursor.Encode());
+      auto ucmp = vstorage->InternalComparator()->user_comparator();
+      // May split output files according to the cursor if it in the user-key
+      // range
+      if (ucmp->CompareWithoutTimestamp(cursor_user_key, smallest_user_key_) >
+              0 &&
+          ucmp->CompareWithoutTimestamp(cursor_user_key, largest_user_key_) <=
+              0) {
+        temp_split_key = cursor;
+      }
+    }
+  }
+  output_split_key_ = temp_split_key;
 }
 
 Compaction::~Compaction() {

package/deps/rocksdb/rocksdb/db/compaction/compaction.h

@@ -177,6 +177,12 @@ class Compaction {
   // moving a single input file to the next level (no merging or splitting)
   bool IsTrivialMove() const;
 
+  // The split user key in the output level if this compaction is required to
+  // split the output files according to the existing cursor in the output
+  // level under round-robin compaction policy. Empty indicates no required
+  // splitting key
+  const InternalKey GetOutputSplitKey() const { return output_split_key_; }
+
   // If true, then the compaction can be done by simply deleting input files.
   bool deletion_compaction() const { return deletion_compaction_; }
 
@@ -379,6 +385,8 @@ class Compaction {
   Temperature output_temperature_;
   // If true, then the compaction can be done by simply deleting input files.
   const bool deletion_compaction_;
+  // should it split the output file using the compact cursor?
+  InternalKey output_split_key_;
 
   // Compaction input files organized by level. Constant after construction
   const std::vector<CompactionInputFiles> inputs_;

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

@@ -197,6 +197,7 @@ void CompactionIterator::Next() {
 
 bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
                                               Slice* skip_until) {
+  // TODO: support compaction filter for wide-column entities
   if (!compaction_filter_ ||
       (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex)) {
     return true;
@@ -519,7 +520,8 @@ void CompactionIterator::NextFromInput() {
       // In the previous iteration we encountered a single delete that we could
       // not compact out.  We will keep this Put, but can drop it's data.
       // (See Optimization 3, below.)
-      if (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex) {
+      if (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex &&
+          ikey_.type != kTypeWideColumnEntity) {
         ROCKS_LOG_FATAL(info_log_, "Unexpected key %s for compaction output",
                         ikey_.DebugString(allow_data_in_errors_, true).c_str());
         assert(false);
@@ -533,7 +535,7 @@ void CompactionIterator::NextFromInput() {
         assert(false);
       }
 
-      if (ikey_.type == kTypeBlobIndex) {
+      if (ikey_.type == kTypeBlobIndex || ikey_.type == kTypeWideColumnEntity) {
         ikey_.type = kTypeValue;
         current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
       }
@@ -689,7 +691,8 @@ void CompactionIterator::NextFromInput() {
             // either way. We will maintain counts of how many mismatches
             // happened
             if (next_ikey.type != kTypeValue &&
-                next_ikey.type != kTypeBlobIndex) {
+                next_ikey.type != kTypeBlobIndex &&
+                next_ikey.type != kTypeWideColumnEntity) {
               ++iter_stats_.num_single_del_mismatch;
             }
 

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

@@ -981,6 +981,21 @@ TEST_F(CompactionIteratorWithSnapshotCheckerTest,
           2 /*earliest_write_conflict_snapshot*/);
 }
 
+// Same as above but with a wide-column entity. In addition to the value getting
+// trimmed, the type of the KV is changed to kTypeValue.
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       KeepSingleDeletionForWriteConflictChecking_WideColumnEntity) {
+  AddSnapshot(2, 0);
+  RunTest({test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeWideColumnEntity)},
+          {"", "fake_entity"},
+          {test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", ""}, 2 /* last_committed_seq */, nullptr /* merge_operator */,
+          nullptr /* compaction_filter */, false /* bottommost_level */,
+          2 /* earliest_write_conflict_snapshot */);
+}
+
 // Compaction filter should keep uncommitted key as-is, and
 //   * Convert the latest value to deletion, and/or
 //   * if latest value is a merge, apply filter to all subsequent merges.

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

@@ -191,7 +191,7 @@ struct CompactionJob::SubcompactionState {
   // The number of bytes overlapping between the current output and
   // grandparent files used in ShouldStopBefore().
   uint64_t overlapped_bytes = 0;
-  // A flag determine whether the key has been seen in ShouldStopBefore()
+  // A flag determines whether the key has been seen in ShouldStopBefore()
   bool seen_key = false;
   // sub compaction job id, which is used to identify different sub-compaction
   // within the same compaction job.
@@ -201,6 +201,9 @@ struct CompactionJob::SubcompactionState {
   // sub-compaction begin.
   bool notify_on_subcompaction_completion = false;
 
+  // A flag determines if this subcompaction has been split by the cursor
+  bool is_split = false;
+
   SubcompactionState(Compaction* c, Slice* _start, Slice* _end, uint64_t size,
                      uint32_t _sub_job_id)
       : compaction(c),
@@ -234,6 +237,23 @@ struct CompactionJob::SubcompactionState {
         &compaction->column_family_data()->internal_comparator();
     const std::vector<FileMetaData*>& grandparents = compaction->grandparents();
 
+    const InternalKey output_split_key = compaction->GetOutputSplitKey();
+    if (output_split_key.Valid() && !is_split) {
+      // Invalid output_split_key indicates that we do not need to split
+      if ((end == nullptr || icmp->user_comparator()->Compare(
+                                 ExtractUserKey(output_split_key.Encode()),
+                                 ExtractUserKey(*end)) < 0) &&
+          (start == nullptr || icmp->user_comparator()->Compare(
+                                   ExtractUserKey(output_split_key.Encode()),
+                                   ExtractUserKey(*start)) > 0)) {
+        // We may only split the output when the cursor is in the range. Split
+        // occurs when the next key is larger than/equal to the cursor
+        if (icmp->Compare(internal_key, output_split_key.Encode()) >= 0) {
+          is_split = true;
+          return true;
+        }
+      }
+    }
     bool grandparant_file_switched = false;
     // Scan to find earliest grandparent file that contains key.
     while (grandparent_index < grandparents.size() &&
@@ -621,6 +641,16 @@ void CompactionJob::GenSubcompactionBoundaries() {
     }
   }
 
+  Slice output_split_user_key;
+  const InternalKey output_split_key = c->GetOutputSplitKey();
+  if (output_split_key.Valid()) {
+    output_split_user_key = ExtractUserKey(output_split_key.Encode());
+    bounds.emplace_back(output_split_key.Encode());
+  } else {
+    // Empty user key indicates that splitting is not required here
+    output_split_user_key = Slice();
+  }
+
   std::sort(bounds.begin(), bounds.end(),
             [cfd_comparator](const Slice& a, const Slice& b) -> bool {
               return cfd_comparator->Compare(ExtractUserKey(a),
@@ -691,7 +721,10 @@ void CompactionJob::GenSubcompactionBoundaries() {
         // need to put an end boundary
         continue;
       }
-      if (sum >= mean) {
+      if (sum >= mean ||
+          (!output_split_user_key.empty() &&
+           cfd_comparator->Compare(ExtractUserKey(ranges[i].range.limit),
+                                   output_split_user_key) == 0)) {
         boundaries_.emplace_back(ExtractUserKey(ranges[i].range.limit));
         sizes_.emplace_back(sum);
         subcompactions--;

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

@@ -1349,6 +1349,61 @@ TEST_F(CompactionPickerTest, CompactionPriRoundRobin) {
   }
 }
 
+TEST_F(CompactionPickerTest, CompactionPriMinOverlappingManyFiles) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  ioptions_.compaction_pri = kMinOverlappingRatio;
+  mutable_cf_options_.max_bytes_for_level_base = 15000000;
+  mutable_cf_options_.max_bytes_for_level_multiplier = 10;
+
+  // file 7 and 8 over lap with the same file, but file 8 is smaller so
+  // it will be picked.
+  Add(2, 13U, "010", "011",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 14U, "020", "021",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 15U, "030", "031",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 16U, "040", "041",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 17U, "050", "051",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 18U, "060", "061",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 19U, "070", "071",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 20U, "080", "081",
+      6100U);  // Overlaps with a large file. Not picked
+
+  Add(2, 6U, "150", "167", 60000000U);  // Overlaps with file 26, 27
+  Add(2, 7U, "168", "169", 60000000U);  // Overlaps with file 27
+  Add(2, 8U, "201", "300", 61000000U);  // Overlaps with file 28, but the file
+                                        // itself is larger. Should be picked.
+  Add(2, 9U, "610", "611",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 10U, "620", "621",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 11U, "630", "631",
+      6100U);  // Overlaps with a large file. Not picked
+  Add(2, 12U, "640", "641",
+      6100U);  // Overlaps with a large file. Not picked
+
+  Add(3, 31U, "001", "100", 260000000U);
+  Add(3, 26U, "160", "165", 260000000U);
+  Add(3, 27U, "166", "170", 260000000U);
+  Add(3, 28U, "180", "400", 260000000U);
+  Add(3, 29U, "401", "500", 260000000U);
+  Add(3, 30U, "601", "700", 260000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Picking file 8 because overlapping ratio is the biggest.
+  ASSERT_EQ(8U, compaction->input(0, 0)->fd.GetNumber());
+}
+
 // This test exhibits the bug where we don't properly reset parent_index in
 // PickCompaction()
 TEST_F(CompactionPickerTest, ParentIndexResetBug) {