Invalidate query compilation cache entries with outdated VIEWs (#1960)

In this PR we add the following algorithm for invalidating cache entries for outdated VIEWs:

1. Store path ids and schema versions of the views that were used in the query in the cache entries, so they can be accessed later.
2. Whenever we retrieve a compilation result from cache, send a request for SchemeCache to check if the schema version of the views used in this query (if any) has not changed since we compiled this query.
3. Send a recompilation request if any view is outdated.

There are two important things to note about this solution:

- We make a SchemeCache request for each repeated query and there is a lot of these in an OLTP-focused database like YDB. However, we have already been sending these request for preliminary (this is not the last check of schema version mismatch (at least for tables)) cache invalidation for tables, so views should not incur an additional performance impact here.
- This solution does not guarantee strong consistency for queries using views, because query cache invalidation will not happen instantly after the view definition is updated. The node should get an update from the SchemeCache, which takes some time.

Author: jepett0

ydb/core/kqp/provider/yql_kikimr_datasource.cpp

@@ -351,6 +351,18 @@ class TKiSourceLoadTableMetadataTransformer : public TGraphTransformerBase {
                 if (!AddCluster(table, res, input, ctx)) {
                     return TStatus::Error;
                 }
+
+                if (const auto& preparingQuery = SessionCtx->Query().PreparingQuery;
+                        preparingQuery
+                        && res.Metadata->Kind == EKikimrTableKind::View
+                ) {
+                    const auto& viewMetadata = *res.Metadata;
+                    auto* viewInfo = preparingQuery->MutablePhysicalQuery()->MutableViewInfos()->Add();
+                    auto* pathId = viewInfo->MutableTableId();
+                    pathId->SetOwnerId(viewMetadata.PathId.OwnerId());
+                    pathId->SetTableId(viewMetadata.PathId.TableId());
+                    viewInfo->SetSchemaVersion(viewMetadata.SchemaVersion);
+                }
             } else {
                 TIssueScopeGuard issueScope(ctx.IssueManager, [input, &table, &ctx]() {
                     return MakeIntrusive<TIssue>(TIssue(ctx.GetPosition(input->Pos()), TStringBuilder()

ydb/core/kqp/session_actor/kqp_query_state.cpp

@@ -71,13 +71,24 @@ bool TKqpQueryState::EnsureTableVersions(const TEvTxProxySchemeCache::TEvNavigat
     return true;
 }
 
+void TKqpQueryState::FillViews(const google::protobuf::RepeatedPtrField< ::NKqpProto::TKqpTableInfo>& views) {
+    for (const auto& view : views) {
+        const auto& pathId = view.GetTableId();
+        const auto schemaVersion = view.GetSchemaVersion();
+        auto [it, isInserted] = TableVersions.emplace(TTableId(pathId.GetOwnerId(), pathId.GetTableId()), schemaVersion);
+        if (!isInserted) {
+            Y_ENSURE(it->second == schemaVersion);
+        }
+    }
+}
 
 std::unique_ptr<TEvTxProxySchemeCache::TEvNavigateKeySet> TKqpQueryState::BuildNavigateKeySet() {
     TableVersions.clear();
 
     for (const auto& tx : PreparedQuery->GetPhysicalQuery().GetTransactions()) {
         FillTables(tx);
     }
+    FillViews(PreparedQuery->GetPhysicalQuery().GetViewInfos());
 
     auto navigate = MakeHolder<NSchemeCache::TSchemeCacheNavigate>();
     navigate->DatabaseName = Database;

ydb/core/kqp/session_actor/kqp_query_state.h

@@ -234,6 +234,8 @@ class TKqpQueryState : public TNonCopyable {
         }
     }
 
+    void FillViews(const google::protobuf::RepeatedPtrField< ::NKqpProto::TKqpTableInfo>& views);
+
     bool NeedCheckTableVersions() const {
         return CompileStats.FromCache;
     }

ydb/core/kqp/ut/view/view_ut.cpp

@@ -1,13 +1,15 @@
 #include <ydb/core/kqp/ut/common/kqp_ut_common.h>
 #include <ydb/library/yql/sql/sql.h>
 #include <ydb/library/yql/utils/log/log.h>
+#include <ydb/public/sdk/cpp/client/ydb_proto/accessor.h>
 
 #include <util/folder/filelist.h>
 
 #include <format>
 
 using namespace NKikimr;
 using namespace NKikimr::NKqp;
+using namespace NYdb;
 using namespace NYdb::NTable;
 
 namespace {
@@ -376,4 +378,51 @@ Y_UNIT_TEST_SUITE(TSelectFromViewTest) {
             ExecuteDataDefinitionQuery(session, ReadWholeFile(pathPrefix + "drop_view.sql"));
         }
     }
+
+    Y_UNIT_TEST(QueryCacheIsUpdated) {
+        TKikimrRunner kikimr(TKikimrSettings().SetWithSampleTables(false));
+        EnableViewsFeatureFlag(kikimr);
+        auto session = kikimr.GetTableClient().CreateSession().GetValueSync().GetSession();
+
+        constexpr const char* viewName = "TheView";
+
+        const auto getCreationQuery = [&viewName](const char* innerQuery) -> TString {
+            return std::format(R"(
+                    CREATE VIEW {} WITH (security_invoker = TRUE) AS {};
+                )",
+                viewName,
+                innerQuery
+            );
+        };
+        constexpr const char* firstInnerQuery = "SELECT 1";
+        ExecuteDataDefinitionQuery(session, getCreationQuery(firstInnerQuery));
+
+        const TString selectFromViewQuery = std::format(R"(
+                SELECT * FROM {};
+            )",
+            viewName
+        );
+        TExecDataQuerySettings queryExecutionSettings;
+        queryExecutionSettings.KeepInQueryCache(true);
+        queryExecutionSettings.CollectQueryStats(ECollectQueryStatsMode::Basic);
+        ExecuteDataModificationQuery(session, selectFromViewQuery, queryExecutionSettings);
+        // make sure the server side cache is working by calling the same query twice
+        const auto cachedQueryRawResult = ExecuteDataModificationQuery(session, selectFromViewQuery, queryExecutionSettings);
+        AssertFromCache(cachedQueryRawResult.GetStats(), true);
+        UNIT_ASSERT_VALUES_EQUAL(GetInteger(GetSingleResult(cachedQueryRawResult)), 1);
+
+        // recreate the view with a different query inside
+        ExecuteDataDefinitionQuery(session, std::format(R"(
+                    DROP VIEW {};
+                )",
+                viewName
+            )
+        );
+        constexpr const char* secondInnerQuery = "SELECT 2";
+        ExecuteDataDefinitionQuery(session, getCreationQuery(secondInnerQuery));
+
+        const auto secondCallRawResult = ExecuteDataModificationQuery(session, selectFromViewQuery, queryExecutionSettings);
+        AssertFromCache(secondCallRawResult.GetStats(), false);
+        UNIT_ASSERT_VALUES_EQUAL(GetInteger(GetSingleResult(secondCallRawResult)), 2);
+    }
 }

ydb/core/protos/kqp_physical.proto

@@ -507,4 +507,6 @@ message TKqpPhyQuery {
     bool HasUncommittedChangesRead = 9;
 
     string QueryDiagnostics = 10;
+
+    repeated TKqpTableInfo ViewInfos = 11;
 }