Add datashard interface for data cleanup operation

Author: lex007in

ydb/core/protos/tx_datashard.proto

@@ -2259,3 +2259,18 @@ message TEvInMemoryStateResponse {
     // Will be set for incomplete replies
     optional bytes ContinuationToken = 3;
 }
+
+// Forces cleanup-related operations in datashard tablet:
+// compaction, GC, GC log and other temporary data erasure.
+// After this operation, all deleted data that was not used at the beginning of the operation
+// is guaranteed to be garbage-collected in the blobstorage.
+// Note: deleted data may still be used for a while, e.g. for history for MVCC.
+message TEvForceDataCleanup {
+    optional uint64 DataCleanupGeneration = 1; // increasing sequence
+}
+
+// Returns DataCleanupGeneration of the last successfully completed TEvForceDataCleanup request.
+// Intermediate requests and corresponding TEvForceDataCleanupResult's may be skipped.
+message TEvForceDataCleanupResult {
+    optional uint64 DataCleanupGeneration = 1; // from corresponding request
+}

ydb/core/tx/datashard/datashard.h

@@ -346,6 +346,9 @@ namespace TEvDataShard {
         EvInMemoryStateRequest,
         EvInMemoryStateResponse,
 
+        EvForceDataCleanup,
+        EvForceDataCleanupResult,
+
         EvEnd
     };
 
@@ -1562,6 +1565,24 @@ namespace TEvDataShard {
         }
     };
 
+    struct TEvForceDataCleanup : TEventPB<TEvForceDataCleanup, NKikimrTxDataShard::TEvForceDataCleanup,
+                                          TEvDataShard::EvForceDataCleanup> {
+        TEvForceDataCleanup() = default;
+
+        TEvForceDataCleanup(ui64 dataCleanupGeneration) {
+            Record.SetDataCleanupGeneration(dataCleanupGeneration);
+        }
+    };
+
+    struct TEvForceDataCleanupResult : TEventPB<TEvForceDataCleanupResult, NKikimrTxDataShard::TEvForceDataCleanupResult,
+                                                TEvDataShard::EvForceDataCleanupResult> {
+        TEvForceDataCleanupResult() = default;
+
+        TEvForceDataCleanupResult(ui64 dataCleanupGeneration) {
+            Record.SetDataCleanupGeneration(dataCleanupGeneration);
+        }
+    };
+
     /**
      * This message is used to ask datashard to compact any borrowed parts it has
      * for the specified user table.

ydb/core/tx/datashard/datashard__data_cleanup.cpp

@@ -0,0 +1,11 @@
+#include "datashard_impl.h"
+
+namespace NKikimr::NDataShard {
+
+void TDataShard::Handle(TEvDataShard::TEvForceDataCleanup::TPtr& ev, const TActorContext& ctx) {
+    // TODO: implement
+    Y_UNUSED(ev);
+    Y_UNUSED(ctx);
+}
+
+} // namespace NKikimr::NDataShard

ydb/core/tx/datashard/datashard_impl.h

@@ -1389,6 +1389,8 @@ class TDataShard
 
     void Handle(TEvIncrementalRestoreScan::TEvFinished::TPtr& ev, const TActorContext& ctx);
 
+    void Handle(TEvDataShard::TEvForceDataCleanup::TPtr& ev, const TActorContext& ctx);
+
     void HandleByReplicationSourceOffsetsServer(STATEFN_SIG);
 
     void DoPeriodicTasks(const TActorContext &ctx);
@@ -3211,6 +3213,7 @@ class TDataShard
             HFunc(TEvPrivate::TEvStatisticsScanFinished, Handle);
             HFuncTraced(TEvPrivate::TEvRemoveSchemaSnapshots, Handle);
             HFunc(TEvIncrementalRestoreScan::TEvFinished, Handle);
+            HFunc(TEvDataShard::TEvForceDataCleanup, Handle);
             default:
                 if (!HandleDefaultEvents(ev, SelfId())) {
                     ALOG_WARN(NKikimrServices::TX_DATASHARD, "TDataShard::StateWork unhandled event type: " << ev->GetTypeRewrite() << " event: " << ev->ToString());

ydb/core/tx/datashard/ya.make

@@ -55,6 +55,7 @@ SRCS(
     datashard__compact_borrowed.cpp
     datashard__compaction.cpp
     datashard__conditional_erase_rows.cpp
+    datashard__data_cleanup.cpp
     datashard__engine_host.cpp
     datashard__engine_host.h
     datashard__get_state_tx.cpp