collecting kValue type tombstone

Summary:
In our testing cluster, we found large amount tombstone has been promoted to kValue type from kMerge after reaching the top level of compaction. Since we used to only collecting tombstone in merge operator, those tombstones can never be collected.

This PR addresses the issue by adding a GC step in compaction filter, which is only for kValue type records. Since those record already reached the top of compaction (no earlier data exists) we can safely remove them in compaction filter without worrying old data appears.

This PR also removes an old optimization in cassandra merge operator for single merge operands.  We need to do GC even on a single operand, so the optimation does not make sense anymore.
Closes https://github.com/facebook/rocksdb/pull/2855

Reviewed By: sagar0

Differential Revision: D5806445

Pulled By: wpc

fbshipit-source-id: 6eb25629d4ce917eb5e8b489f64a6aa78c7d270b

java/rocksjni/cassandra_compactionfilterjni.cc

@@ -11,12 +11,13 @@
 /*
  * Class:     org_rocksdb_CassandraCompactionFilter
  * Method:    createNewCassandraCompactionFilter0
- * Signature: ()J
+ * Signature: (ZI)J
  */
 jlong Java_org_rocksdb_CassandraCompactionFilter_createNewCassandraCompactionFilter0(
-    JNIEnv* env, jclass jcls, jboolean purge_ttl_on_expiration) {
-  auto* compaction_filter =
-      new rocksdb::cassandra::CassandraCompactionFilter(purge_ttl_on_expiration);
+    JNIEnv* env, jclass jcls, jboolean purge_ttl_on_expiration,
+    jint gc_grace_period_in_seconds) {
+  auto* compaction_filter = new rocksdb::cassandra::CassandraCompactionFilter(
+      purge_ttl_on_expiration, gc_grace_period_in_seconds);
   // set the native handle to our native compaction filter
   return reinterpret_cast<jlong>(compaction_filter);
 }

java/src/main/java/org/rocksdb/CassandraCompactionFilter.java

@@ -10,9 +10,10 @@ package org.rocksdb;
  */
 public class CassandraCompactionFilter
     extends AbstractCompactionFilter<Slice> {
-  public CassandraCompactionFilter(boolean purgeTtlOnExpiration) {
-      super(createNewCassandraCompactionFilter0(purgeTtlOnExpiration));
+  public CassandraCompactionFilter(boolean purgeTtlOnExpiration, int gcGracePeriodInSeconds) {
+    super(createNewCassandraCompactionFilter0(purgeTtlOnExpiration, gcGracePeriodInSeconds));
   }
 
-  private native static long createNewCassandraCompactionFilter0(boolean purgeTtlOnExpiration);
+  private native static long createNewCassandraCompactionFilter0(
+      boolean purgeTtlOnExpiration, int gcGracePeriodInSeconds);
 }

utilities/cassandra/cassandra_compaction_filter.cc

@@ -32,6 +32,10 @@ CompactionFilter::Decision CassandraCompactionFilter::FilterV2(
           ? row_value.RemoveExpiredColumns(&value_changed)
           : row_value.ConvertExpiredColumnsToTombstones(&value_changed);
 
+  if (value_type == ValueType::kValue) {
+    compacted = compacted.RemoveTombstones(gc_grace_period_in_seconds_);
+  }
+
   if(compacted.Empty()) {
     return Decision::kRemove;
   }

utilities/cassandra/cassandra_compaction_filter.h

@@ -18,22 +18,25 @@ namespace cassandra {
  * tombstones first, then be eventally removed after gc grace period.
  * `purge_ttl_on_expiration` should only be on in the case all the
  * writes have same ttl setting, otherwise it could bring old data back.
+ *
+ * Compaction filter is also in charge of removing tombstone that has been
+ * promoted to kValue type after serials of merging in compaction.
  */
 class CassandraCompactionFilter : public CompactionFilter {
 public:
-  explicit CassandraCompactionFilter(bool purge_ttl_on_expiration)
-    : purge_ttl_on_expiration_(purge_ttl_on_expiration) {}
+ explicit CassandraCompactionFilter(bool purge_ttl_on_expiration,
+                                    int32_t gc_grace_period_in_seconds)
+     : purge_ttl_on_expiration_(purge_ttl_on_expiration),
+       gc_grace_period_in_seconds_(gc_grace_period_in_seconds) {}
 
  const char* Name() const override;
-  virtual Decision FilterV2(int level,
-                            const Slice& key,
-                            ValueType value_type,
-                            const Slice& existing_value,
-                            std::string* new_value,
+ virtual Decision FilterV2(int level, const Slice& key, ValueType value_type,
+                           const Slice& existing_value, std::string* new_value,
                            std::string* skip_until) const override;
 
 private:
   bool purge_ttl_on_expiration_;
+  int32_t gc_grace_period_in_seconds_;
 };
 }  // namespace cassandra
 }  // namespace rocksdb

utilities/cassandra/cassandra_format_test.cc

@@ -122,6 +122,19 @@ TEST(ExpiringColumnTest, ExpiringColumn) {
       == 0);
 }
 
+TEST(TombstoneTest, TombstoneCollectable) {
+  int32_t now = (int32_t)time(nullptr);
+  int32_t gc_grace_seconds = 16440;
+  EXPECT_TRUE(Tombstone(ColumnTypeMask::DELETION_MASK, 0,
+                        now - gc_grace_seconds,
+                        ToMicroSeconds(now - gc_grace_seconds))
+                  .Collectable(gc_grace_seconds));
+  EXPECT_FALSE(Tombstone(ColumnTypeMask::DELETION_MASK, 0,
+                         now - gc_grace_seconds + 1,
+                         ToMicroSeconds(now - gc_grace_seconds + 1))
+                   .Collectable(gc_grace_seconds));
+}
+
 TEST(TombstoneTest, Tombstone) {
   int8_t mask = ColumnTypeMask::DELETION_MASK;
   int8_t index = 2;

utilities/cassandra/cassandra_functional_test.cc

@@ -26,9 +26,7 @@ const std::string kDbName = test::TmpDir() + "/cassandra_functional_test";
 class CassandraStore {
  public:
   explicit CassandraStore(std::shared_ptr<DB> db)
-      : db_(db),
-        merge_option_(),
-        get_option_() {
+      : db_(db), write_option_(), get_option_() {
     assert(db);
   }
 
@@ -36,7 +34,7 @@ class CassandraStore {
     std::string result;
     val.Serialize(&result);
     Slice valSlice(result.data(), result.size());
-    auto s = db_->Merge(merge_option_, key, valSlice);
+    auto s = db_->Merge(write_option_, key, valSlice);
 
     if (s.ok()) {
       return true;
@@ -46,6 +44,19 @@ class CassandraStore {
     }
   }
 
+  bool Put(const std::string& key, const RowValue& val) {
+    std::string result;
+    val.Serialize(&result);
+    Slice valSlice(result.data(), result.size());
+    auto s = db_->Put(write_option_, key, valSlice);
+    if (s.ok()) {
+      return true;
+    } else {
+      std::cerr << "ERROR " << s.ToString() << std::endl;
+      return false;
+    }
+  }
+
   void Flush() {
     dbfull()->TEST_FlushMemTable();
     dbfull()->TEST_WaitForCompact();
@@ -75,21 +86,23 @@ class CassandraStore {
 
  private:
   std::shared_ptr<DB> db_;
-  WriteOptions merge_option_;
+  WriteOptions write_option_;
   ReadOptions get_option_;
 
   DBImpl* dbfull() { return reinterpret_cast<DBImpl*>(db_.get()); }
-
 };
 
 class TestCompactionFilterFactory : public CompactionFilterFactory {
 public:
-  explicit TestCompactionFilterFactory(bool purge_ttl_on_expiration)
-    : purge_ttl_on_expiration_(purge_ttl_on_expiration) {}
+ explicit TestCompactionFilterFactory(bool purge_ttl_on_expiration,
+                                      int32_t gc_grace_period_in_seconds)
+     : purge_ttl_on_expiration_(purge_ttl_on_expiration),
+       gc_grace_period_in_seconds_(gc_grace_period_in_seconds) {}
 
  virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
      const CompactionFilter::Context& context) override {
-    return unique_ptr<CompactionFilter>(new CassandraCompactionFilter(purge_ttl_on_expiration_));
+   return unique_ptr<CompactionFilter>(new CassandraCompactionFilter(
+       purge_ttl_on_expiration_, gc_grace_period_in_seconds_));
   }
 
   virtual const char* Name() const override {
@@ -98,6 +111,7 @@ public:
 
 private:
   bool purge_ttl_on_expiration_;
+  int32_t gc_grace_period_in_seconds_;
 };
 
 
@@ -113,7 +127,8 @@ public:
     Options options;
     options.create_if_missing = true;
     options.merge_operator.reset(new CassandraValueMergeOperator(gc_grace_period_in_seconds_));
-    auto* cf_factory = new TestCompactionFilterFactory(purge_ttl_on_expiration_);
+    auto* cf_factory = new TestCompactionFilterFactory(
+        purge_ttl_on_expiration_, gc_grace_period_in_seconds_);
     options.compaction_filter_factory.reset(cf_factory);
     EXPECT_OK(DB::Open(options, kDbName, &db));
     return std::shared_ptr<DB>(db);
@@ -275,6 +290,19 @@ TEST_F(CassandraFunctionalTest,
   VerifyRowValueColumns(gced.columns_, 0, kColumn, 1, ToMicroSeconds(now));
 }
 
+TEST_F(CassandraFunctionalTest, CompactionShouldRemoveTombstoneFromPut) {
+  purge_ttl_on_expiration_ = true;
+  CassandraStore store(OpenDb());
+  int64_t now = time(nullptr);
+
+  store.Put("k1", CreateTestRowValue({
+    std::make_tuple(kTombstone, 0, ToMicroSeconds(now - gc_grace_period_in_seconds_ - 1)),
+  }));
+
+  store.Flush();
+  store.Compact();
+  ASSERT_FALSE(std::get<0>(store.Get("k1")));
+}
 
 } // namespace cassandra
 } // namespace rocksdb

utilities/cassandra/merge_operator.cc

@@ -22,13 +22,6 @@ bool CassandraValueMergeOperator::FullMergeV2(
     MergeOperationOutput* merge_out) const {
   // Clear the *new_value for writing.
   merge_out->new_value.clear();
-
-  if (merge_in.existing_value == nullptr && merge_in.operand_list.size() == 1) {
-    // Only one operand
-    merge_out->existing_operand = merge_in.operand_list.back();
-    return true;
-  }
-
   std::vector<RowValue> row_values;
   if (merge_in.existing_value) {
     row_values.push_back(