Fix three schema mismatch bugs under disaggregated arch (#10530) (#10579)

ti-chi-bot · gengliqi · web-flow · commit c529adf6ebac · 2025-12-01T06:14:38.000Z
close #10226 Fix three schema mismatch bugs under disaggregated arch Signed-off-by: gengliqi <gengliqiii@gmail.com> Co-authored-by: Liqi Geng <gengliqiii@gmail.com>
diff --git a/dbms/src/Flash/Coprocessor/GenSchemaAndColumn.cpp b/dbms/src/Flash/Coprocessor/GenSchemaAndColumn.cpp
@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+#include <DataStreams/GeneratedColumnPlaceholderBlockInputStream.h>
 #include <Flash/Coprocessor/GenSchemaAndColumn.h>
 #include <Storages/DeltaMerge/DeltaMergeDefines.h>
 #include <Storages/MutableSupport.h>
@@ -94,14 +95,23 @@ NamesAndTypes genNamesAndTypes(const TiDBTableScan & table_scan, const StringRef
     return genNamesAndTypes(table_scan.getColumns(), column_prefix);
 }
 
-std::tuple<DM::ColumnDefinesPtr, int> genColumnDefinesForDisaggregatedRead(const TiDBTableScan & table_scan)
+std::tuple<DM::ColumnDefinesPtr, int, std::vector<std::tuple<UInt64, String, DataTypePtr>>> genColumnDefinesForDisaggregatedRead(
+    const TiDBTableScan & table_scan)
 {
     auto column_defines = std::make_shared<DM::ColumnDefines>();
     int extra_table_id_index = MutSup::invalid_col_id;
     column_defines->reserve(table_scan.getColumnSize());
+    std::vector<std::tuple<UInt64, String, DataTypePtr>> generated_column_infos;
     for (Int32 i = 0; i < table_scan.getColumnSize(); ++i)
     {
         const auto & column_info = table_scan.getColumns()[i];
+        if (column_info.hasGeneratedColumnFlag())
+        {
+            const auto & data_type = getDataTypeByColumnInfoForComputingLayer(column_info);
+            const auto & col_name = GeneratedColumnPlaceholderBlockInputStream::getColumnName(i);
+            generated_column_infos.push_back(std::make_tuple(i, col_name, data_type));
+            continue;
+        }
         // Now the upper level seems treat disagg read as an ExchangeReceiver output, so
         // use this as output column prefix.
         // Even if the id is pk_column or extra_table_id, we still output it as
@@ -117,10 +127,6 @@ std::tuple<DM::ColumnDefinesPtr, int> genColumnDefinesForDisaggregatedRead(const
             break;
         case MutSup::extra_table_id_col_id:
         {
-            column_defines->emplace_back(DM::ColumnDefine{
-                MutSup::extra_table_id_col_id,
-                output_name, // MutSup::extra_table_id_column_name
-                MutSup::getExtraTableIdColumnType()});
             extra_table_id_index = i;
             break;
         }
@@ -133,7 +139,7 @@ std::tuple<DM::ColumnDefinesPtr, int> genColumnDefinesForDisaggregatedRead(const
             break;
         }
     }
-    return {std::move(column_defines), extra_table_id_index};
+    return {std::move(column_defines), extra_table_id_index, std::move(generated_column_infos)};
 }
 
 ColumnsWithTypeAndName getColumnWithTypeAndName(const NamesAndTypes & names_and_types)
diff --git a/dbms/src/Flash/Coprocessor/GenSchemaAndColumn.h b/dbms/src/Flash/Coprocessor/GenSchemaAndColumn.h
@@ -34,7 +34,8 @@ NamesAndTypes genNamesAndTypes(const TiDB::ColumnInfos & column_infos, const Str
 ColumnsWithTypeAndName getColumnWithTypeAndName(const NamesAndTypes & names_and_types);
 NamesAndTypes toNamesAndTypes(const DAGSchema & dag_schema);
 
-// The column defines and `extra table id index`
-std::tuple<DM::ColumnDefinesPtr, int> genColumnDefinesForDisaggregatedRead(const TiDBTableScan & table_scan);
+// The column defines, `extra table id index` and `generated columns info` for disaggregated read.
+std::tuple<DM::ColumnDefinesPtr, int, std::vector<std::tuple<UInt64, String, DataTypePtr>>> genColumnDefinesForDisaggregatedRead(
+    const TiDBTableScan & table_scan);
 
 } // namespace DB
diff --git a/dbms/src/Storages/StorageDisaggregated.h b/dbms/src/Storages/StorageDisaggregated.h
@@ -108,7 +108,7 @@ class StorageDisaggregated : public IStorage
     std::tuple<DM::RSOperatorPtr, DM::ColumnRangePtr> buildRSOperatorAndColumnRange(
         const Context & db_context,
         const DM::ColumnDefinesPtr & columns_to_read);
-    std::variant<DM::Remote::RNWorkersPtr, DM::SegmentReadTaskPoolPtr> packSegmentReadTasks(
+    std::tuple<std::variant<DM::Remote::RNWorkersPtr, DM::SegmentReadTaskPoolPtr>, DM::ColumnDefinesPtr> packSegmentReadTasks(
         const Context & db_context,
         DM::SegmentReadTasks && read_tasks,
         const DM::ColumnDefinesPtr & column_defines,
@@ -161,5 +161,7 @@ class StorageDisaggregated : public IStorage
     std::unique_ptr<DAGExpressionAnalyzer> analyzer;
     static constexpr auto ZONE_LABEL_KEY = "zone";
     std::optional<String> zone_label;
+    // For generated column, just need a placeholder, and TiDB will fill this column.
+    std::vector<std::tuple<UInt64, String, DataTypePtr>> generated_column_infos;
 };
 } // namespace DB
diff --git a/dbms/src/Storages/StorageDisaggregatedRemote.cpp b/dbms/src/Storages/StorageDisaggregatedRemote.cpp
@@ -115,6 +115,8 @@ BlockInputStreams StorageDisaggregated::readThroughS3(const Context & db_context
         num_streams,
         pipeline,
         scan_context);
+    // handle generated column if necessary.
+    executeGeneratedColumnPlaceholder(generated_column_infos, log, pipeline);
 
     NamesAndTypes source_columns;
     source_columns.reserve(table_scan.getColumnSize());
@@ -149,6 +151,8 @@ void StorageDisaggregated::readThroughS3(
         buildReadTaskWithBackoff(db_context, scan_context),
         num_streams,
         scan_context);
+    // handle generated column if necessary.
+    executeGeneratedColumnPlaceholder(exec_context, group_builder, generated_column_infos, log);
 
     NamesAndTypes source_columns;
     auto header = group_builder.getCurrentHeader();
@@ -609,13 +613,14 @@ std::tuple<DM::RSOperatorPtr, DM::ColumnRangePtr> StorageDisaggregated::buildRSO
     return {rs_operator, column_range};
 }
 
-std::variant<DM::Remote::RNWorkersPtr, DM::SegmentReadTaskPoolPtr> StorageDisaggregated::packSegmentReadTasks(
-    const Context & db_context,
-    DM::SegmentReadTasks && read_tasks,
-    const DM::ColumnDefinesPtr & column_defines,
-    const DM::ScanContextPtr & scan_context,
-    size_t num_streams,
-    int extra_table_id_index)
+std::tuple<std::variant<DM::Remote::RNWorkersPtr, DM::SegmentReadTaskPoolPtr>, DM::ColumnDefinesPtr> StorageDisaggregated::
+    packSegmentReadTasks(
+        const Context & db_context,
+        DM::SegmentReadTasks && read_tasks,
+        const DM::ColumnDefinesPtr & column_defines,
+        const DM::ScanContextPtr & scan_context,
+        size_t num_streams,
+        int extra_table_id_index)
 {
     const auto & executor_id = table_scan.getTableScanExecutorID();
 
@@ -651,53 +656,61 @@ std::variant<DM::Remote::RNWorkersPtr, DM::SegmentReadTaskPoolPtr> StorageDisagg
     scan_context->read_mode = read_mode;
     const UInt64 start_ts = sender_target_mpp_task_id.gather_id.query_id.start_ts;
     const auto enable_read_thread = db_context.getSettingsRef().dt_enable_read_thread;
+    const auto & final_columns_defines = push_down_executor && push_down_executor->extra_cast
+        ? push_down_executor->columns_after_cast
+        : column_defines;
     RUNTIME_CHECK(num_streams > 0, num_streams);
     LOG_INFO(
         log,
         "packSegmentReadTasks: enable_read_thread={} read_mode={} is_fast_scan={} keep_order={} task_count={} "
-        "num_streams={} column_defines={}",
+        "num_streams={} column_defines={} final_columns_defines={}",
         enable_read_thread,
         magic_enum::enum_name(read_mode),
         table_scan.isFastScan(),
         table_scan.keepOrder(),
         read_tasks.size(),
         num_streams,
-        *column_defines);
+        *column_defines,
+        *final_columns_defines);
 
     if (enable_read_thread)
     {
         // Under disagg arch, now we use blocking IO to read data from cloud storage. So it require more active
         // segments to fully utilize the read threads.
         const size_t read_thread_num_active_seg = 10 * num_streams;
-        return std::make_shared<DM::SegmentReadTaskPool>(
-            extra_table_id_index,
-            *column_defines,
-            push_down_executor,
-            start_ts,
-            db_context.getSettingsRef().max_block_size,
-            read_mode,
-            std::move(read_tasks),
-            /*after_segment_read*/ [](const DM::DMContextPtr &, const DM::SegmentPtr &) {},
-            log->identifier(),
-            /*enable_read_thread*/ true,
-            num_streams,
-            read_thread_num_active_seg,
-            context.getDAGContext()->getKeyspaceID(),
-            context.getDAGContext()->getResourceGroupName());
+        return {
+            std::make_shared<DM::SegmentReadTaskPool>(
+                extra_table_id_index,
+                *final_columns_defines,
+                push_down_executor,
+                start_ts,
+                db_context.getSettingsRef().max_block_size,
+                read_mode,
+                std::move(read_tasks),
+                /*after_segment_read*/ [](const DM::DMContextPtr &, const DM::SegmentPtr &) {},
+                log->identifier(),
+                /*enable_read_thread*/ true,
+                num_streams,
+                read_thread_num_active_seg,
+                context.getDAGContext()->getKeyspaceID(),
+                context.getDAGContext()->getResourceGroupName()),
+            final_columns_defines};
     }
     else
     {
-        return DM::Remote::RNWorkers::create(
-            db_context,
-            std::move(read_tasks),
-            {
-                .log = log->getChild(executor_id),
-                .columns_to_read = column_defines,
-                .start_ts = start_ts,
-                .push_down_executor = push_down_executor,
-                .read_mode = read_mode,
-            },
-            num_streams);
+        return {
+            DM::Remote::RNWorkers::create(
+                db_context,
+                std::move(read_tasks),
+                {
+                    .log = log->getChild(executor_id),
+                    .columns_to_read = final_columns_defines,
+                    .start_ts = start_ts,
+                    .push_down_executor = push_down_executor,
+                    .read_mode = read_mode,
+                },
+                num_streams),
+            final_columns_defines};
     }
 }
 
@@ -739,8 +752,11 @@ void StorageDisaggregated::buildRemoteSegmentInputStreams(
     const DM::ScanContextPtr & scan_context)
 {
     // Build the input streams to read blocks from remote segments
-    auto [column_defines, extra_table_id_index] = genColumnDefinesForDisaggregatedRead(table_scan);
-    auto packed_read_tasks = packSegmentReadTasks(
+    DM::ColumnDefinesPtr column_defines;
+    int extra_table_id_index;
+    std::tie(column_defines, extra_table_id_index, generated_column_infos)
+        = genColumnDefinesForDisaggregatedRead(table_scan);
+    auto [packed_read_tasks, final_column_defines] = packSegmentReadTasks(
         db_context,
         std::move(read_tasks),
         column_defines,
@@ -751,7 +767,7 @@ void StorageDisaggregated::buildRemoteSegmentInputStreams(
 
     InputStreamBuilder builder{
         .tracing_id = log->identifier(),
-        .columns_to_read = column_defines,
+        .columns_to_read = final_column_defines,
         .extra_table_id_index = extra_table_id_index,
     };
     for (size_t stream_idx = 0; stream_idx < num_streams; ++stream_idx)
@@ -810,8 +826,11 @@ void StorageDisaggregated::buildRemoteSegmentSourceOps(
     const DM::ScanContextPtr & scan_context)
 {
     // Build the input streams to read blocks from remote segments
-    auto [column_defines, extra_table_id_index] = genColumnDefinesForDisaggregatedRead(table_scan);
-    auto packed_read_tasks = packSegmentReadTasks(
+    DM::ColumnDefinesPtr column_defines;
+    int extra_table_id_index;
+    std::tie(column_defines, extra_table_id_index, generated_column_infos)
+        = genColumnDefinesForDisaggregatedRead(table_scan);
+    auto [packed_read_tasks, final_column_defines] = packSegmentReadTasks(
         db_context,
         std::move(read_tasks),
         column_defines,
@@ -821,7 +840,7 @@ void StorageDisaggregated::buildRemoteSegmentSourceOps(
 
     SourceOpBuilder builder{
         .tracing_id = log->identifier(),
-        .column_defines = column_defines,
+        .column_defines = final_column_defines,
         .extra_table_id_index = extra_table_id_index,
         .exec_context = exec_context,
     };
diff --git a/tests/docker/next-gen-config/tidb.toml b/tests/docker/next-gen-config/tidb.toml
@@ -19,6 +19,8 @@ disaggregated-tiflash = true
 # Now tests are ran on the SYSTEM keyspace tidb.
 keyspace-name = "SYSTEM"
 
+tikv-worker-url = "tikv-worker0:19000"
+
 enable-telemetry = false
 temp-dir = "/data/tmp"
 [performance]
diff --git a/tests/docker/next-gen-config/tikv-worker.toml b/tests/docker/next-gen-config/tikv-worker.toml
@@ -0,0 +1,28 @@
+# Copyright 2025 PingCAP, Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# object storage for next-gen
+[dfs]
+prefix = "tikv"
+s3-endpoint = "http://minio0:9000"
+s3-key-id = "minioadmin"
+s3-secret-key = "minioadmin"
+s3-bucket = "tiflash-test"
+s3-region = "local"
+
+[schema-manager]
+dir = "/data/schemas"
+enabled = true
+keyspace-refresh-interval = "10s"
+schema-refresh-threshold = 1
diff --git a/tests/docker/next-gen-config/tikv.toml b/tests/docker/next-gen-config/tikv.toml
@@ -18,6 +18,7 @@ reserve-space = "0"
 # Enable keyspace and ttl for next-gen
 api-version = 2
 enable-ttl = true
+low-space-threshold = 0
 
 [raftstore]
 capacity = "100GB"
diff --git a/tests/docker/next-gen-yaml/cluster.yaml b/tests/docker/next-gen-yaml/cluster.yaml
@@ -49,6 +49,19 @@ services:
       - "pd0"
       - "minio0"
     restart: on-failure
+  tikv-worker0:
+    image: ${TIKV_IMAGE:-us-docker.pkg.dev/pingcap-testing-account/hub/tikv/tikv/image:dedicated-next-gen}
+    security_opt:
+      - seccomp:unconfined
+    volumes:
+      - ./next-gen-config/tikv-worker.toml:/tikv-worker.toml:ro
+      - ./data/tikv-worker0:/data
+      - ./log/tikv-worker0:/log
+    entrypoint: /tikv-worker
+    command: --addr=0.0.0.0:19000 --pd-endpoints=pd0:2379 --config=/tikv-worker.toml --data-dir=/data --log-file=/log/tikv-worker.log
+    depends_on:
+      - "pd0"
+    restart: on-failure
   tidb0:
     image: ${TIDB_IMAGE:-us-docker.pkg.dev/pingcap-testing-account/hub/pingcap/tidb/images/tidb-server:master-next-gen}
     security_opt:
diff --git a/tests/fullstack-test-next-gen/run.sh b/tests/fullstack-test-next-gen/run.sh
@@ -69,15 +69,15 @@ if [[ -n "$ENABLE_NEXT_GEN" && "$ENABLE_NEXT_GEN" != "false" && "$ENABLE_NEXT_GE
     ENV_ARGS="ENABLE_NEXT_GEN=true verbose=${verbose} "
     # most failpoints are expected to be set on the compute layer, use tiflash-cn0 to run tests
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test/sample.test"
-    ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test-index/vector"
+    ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test-index"
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test-next-gen/placement"
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test2/clustered_index"
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test2/dml"
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test2/variables"
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test2/mpp"
-    # TODO: enable the following tests after they are fixed. And maybe we need to split them into parallel pipelines because they take too long to run.
-    #${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test/expr"
-    #${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test/mpp"
+    # maybe we need to split them into parallel pipelines because they take too long to run.
+    ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test/expr"
+    ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" exec -T tiflash-cn0 bash -c "cd /tests ; ${ENV_ARGS} ./run-test.sh fullstack-test/mpp"
     ${COMPOSE} -f next-gen-cluster.yaml -f "${DISAGG_TIFLASH_YAML}" down
     clean_data_log
 
diff --git a/tests/fullstack-test/mpp/extra_physical_table_column.test b/tests/fullstack-test/mpp/extra_physical_table_column.test
@@ -12,9 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# line 27: Pessimistic lock response corrupted
-#SKIP_FOR_NEXT_GEN
-
 # Preparation.
 => DBGInvoke __init_fail_point()
 
