Cherry pick v3.2.0 (0627-0630) (#4370)

* enhance nightly tck test job (#4355) * add lock when multithread access shared variable (#4360) * Implement inner parallel procesing by using Scatter/Gather model. (#4361) * Multi job traverse. Some debug infos. Multi job AppendVertices handle resp. Multi job Project. Enhance AppendVertices. Multi job filter. Fix row cnt and delete debug info. Fix AppendV. Add debug info. Rm debug info. Fix copy in traverse. Set max_job_size to default 20. Add runMultiJobs template. Impl Project/Filter/AppendV. Impl Join. Fix double free for join. Fix filter on GetNeighborsIter. Fix prop iter. Fix filter on prop iter. Fix left join by clearing hash tabls manually. Report error in filter. Fix tests: the order is not guarentted in scatter gather model. Iterates to begin out of job. Modify the default config. Fix appendv. Support run multi job mode test. Test multi job. Rebase and compatible with single job. * Update the logs and flags. * Add comment. * Fix nightly. * Fix test. * Fix compile. * Fix test. * Address comment and fix test. Co-authored-by: Sophie <[email protected]> * Async gc. (#4362) * Add GC. * Impl gc. * Fix linking. * Using thread pool. * Using try_dequeue. * Stop workers when destruct. * Set default worker size to 8. * Add repeat task for all. * Update flags. * Add comments. * Fix nightly. Co-authored-by: Harris.Chu <[email protected]> Co-authored-by: jimingquan <[email protected]> Co-authored-by: cpw <[email protected]>
vesoft-inc · Jun 30, 2022 · 518c7d7 · 518c7d7
1 parent dfe737c
commit 518c7d7
Show file tree

Hide file tree

Showing 44 changed files with 1,029 additions and 61 deletions.
diff --git a/.github/workflows/nightly.yml b/.github/workflows/nightly.yml
@@ -2,7 +2,7 @@ name: nightly
 
 on:
   schedule:
-    - cron: '0 18 * * *'
+    - cron: "0 18 * * *"
 
 concurrency:
   group: nightly
@@ -35,7 +35,7 @@ jobs:
       - uses: actions/checkout@v2
       - uses: actions/setup-go@v2
         with:
-          go-version: '^1.16.7'
+          go-version: "^1.16.7"
       - name: package
         run: ./package/package.sh -t Release -r ON -p OFF -s FALSE -k OFF
       - name: output some vars
@@ -199,4 +199,99 @@ jobs:
         if: ${{ failure() }}
         with:
           name: ${{ matrix.os }}-${{ matrix.compiler }}-nebula-test-logs
-          path:  ./build/server_*/logs/
+          path: ./build/server_*/logs/
+
+  test:
+    name: Tck test
+    needs: package
+    runs-on: [self-hosted, nebula]
+    strategy:
+      fail-fast: false
+      matrix:
+        os:
+          - ubuntu1804
+          - ubuntu2004
+          - centos7
+          - centos8
+        extra_config:
+          - "ENABLE_SSL=true CA_SIGNED=true QUERY_CONCURRENTLY=false"
+          - "ENABLE_SSL=false CA_SIGNED=false QUERY_CONCURRENTLY=true"
+    env:
+      OSS_DIR: nebula-graph/package/nightly
+    container:
+      image: vesoft/nebula-dev:${{ matrix.os }}
+    steps:
+      - uses: webiny/[email protected]
+        with:
+          run: sh -c "find . -mindepth 1 -delete"
+      - uses: actions/checkout@v2
+      - name: output some vars
+        id: vars
+        env:
+          SHA_EXT: sha256sum.txt
+        run: |
+          subdir=$(date -u +%Y.%m.%d -d "1 days ago")
+          echo "::set-output name=subdir::$subdir"
+
+      - id: oss_package
+        run: |
+          case ${{ matrix.os }} in
+          ubuntu1804)
+            echo "::set-output name=p::nebula-graph-${{ steps.vars.outputs.subdir }}-nightly.ubuntu1804.amd64.tar.gz"
+            ;;
+          ubuntu2004)
+            echo "::set-output name=p::nebula-graph-${{ steps.vars.outputs.subdir }}-nightly.ubuntu2004.amd64.tar.gz"
+            ;;
+          centos7)
+            echo "::set-output name=p::nebula-graph-${{ steps.vars.outputs.subdir }}-nightly.el7.x86_64.tar.gz"
+            ;;
+          centos8)
+            echo "::set-output name=p::nebula-graph-${{ steps.vars.outputs.subdir }}-nightly.el8.x86_64.tar.gz"
+            ;;
+          esac
+      - name: Prepare environment
+        id: prepare
+        run: |
+          [ -d build/ ] && rm -rf build/* || mkdir -p build
+          make init -C tests
+      - name: CMake
+        id: cmake
+        run: |
+          echo "::set-output name=j::8"
+      - name: download from oss
+        run: |
+          ossutil64 -e ${{ secrets.OSS_ENDPOINT }} \
+                    -i ${{ secrets.OSS_ID }} \
+                    -k  ${{ secrets.OSS_SECRET }} \
+                    -f cp oss://${{ env.OSS_DIR  }}/${{ steps.vars.outputs.subdir }}/${{ steps.oss_package.outputs.p }} \
+                    build/.
+          tar zxvf build/${{ steps.oss_package.outputs.p }} -C build
+          d=`echo ${{ steps.oss_package.outputs.p }} | sed 's/.tar.gz//'`
+          mv build/${d}/* build/.
+
+      - name: Setup cluster
+        run: |
+          make CONTAINERIZED=true ${{ matrix.extra_config }} up
+        working-directory: tests/
+        timeout-minutes: 2
+      - name: Pytest
+        run: |
+          make RM_DIR=false DEBUG=false J=${{ steps.cmake.outputs.j }} test
+        working-directory: tests/
+        timeout-minutes: 15
+      - name: TCK
+        run: |
+          make RM_DIR=false DEBUG=false J=${{ steps.cmake.outputs.j }} tck
+        working-directory: tests/
+        timeout-minutes: 60
+      - name: Down cluster
+        run: |
+          make RM_DIR=false down
+        working-directory: tests/
+        timeout-minutes: 2
+      - name: Upload logs
+        uses: actions/upload-artifact@v2
+        if: ${{ failure() }}
+        with:
+          name: ${{ matrix.os }}-${{ matrix.compiler }}-nebula-test-logs
+          path: ./build/server_*/logs*/
diff --git a/.github/workflows/rc.yml b/.github/workflows/rc.yml
@@ -185,6 +185,9 @@ jobs:
           - ubuntu2004
           - centos7
           - centos8
+        extra_config:
+          - "ENABLE_SSL=true CA_SIGNED=true QUERY_CONCURRENTLY=false"
+          - "ENABLE_SSL=false CA_SIGNED=false QUERY_CONCURRENTLY=true"
     env:
       OSS_DIR: nebula-graph/rc
     container:
@@ -235,7 +238,7 @@ jobs:
      
       - name: Setup cluster
         run: |
-          make CONTAINERIZED=true ENABLE_SSL=true CA_SIGNED=true up
+          make CONTAINERIZED=true ${{ matrix.extra_config }} up
         working-directory: tests/
         timeout-minutes: 2
       - name: Pytest
@@ -258,4 +261,4 @@ jobs:
         if: ${{ failure() }}
         with:
           name: ${{ matrix.os }}-${{ matrix.compiler }}-nebula-test-logs
-          path:  ./build/server_*/logs/
+          path:  ./build/server_*/logs*/
diff --git a/src/common/expression/test/CMakeLists.txt b/src/common/expression/test/CMakeLists.txt
@@ -47,6 +47,7 @@ set(expression_test_common_libs
     $<TARGET_OBJECTS:graph_obj>
     $<TARGET_OBJECTS:ssl_obj>
     $<TARGET_OBJECTS:memory_obj>
+    $<TARGET_OBJECTS:gc_obj>
 )
 
 

diff --git a/src/common/thread/GenericThreadPool.h b/src/common/thread/GenericThreadPool.h
@@ -113,6 +113,16 @@ class GenericThreadPool final : public boost::noncopyable, public nebula::cpp::N
    */
   void purgeTimerTask(uint64_t id);
 
+  /**
+   * To add a repeated timer task for all workers which will be executed in each period.
+   * @ms      interval in milliseconds
+   * @task    a callable object
+   * @args    variadic arguments
+   * @return  ID of the added task, unique for this worker
+   */
+  template <typename F, typename... Args>
+  void addRepeatTaskForAll(size_t ms, F &&f, Args &&... args);
+
  private:
   size_t nrThreads_{0};
   std::atomic<size_t> nextThread_{0};
@@ -156,6 +166,12 @@ uint64_t GenericThreadPool::addRepeatTask(size_t ms, F &&f, Args &&... args) {
   return ((idx << GenericWorker::TIMER_ID_BITS) | id);
 }
 
+template <typename F, typename... Args>
+void GenericThreadPool::addRepeatTaskForAll(size_t ms, F &&f, Args &&... args) {
+  for (auto idx = 0UL; idx < nrThreads_; ++idx) {
+    pool_[idx]->addRepeatTask(ms, std::forward<F>(f), std::forward<Args>(args)...);
+  }
+}
 }  // namespace thread
 }  // namespace nebula
 

diff --git a/src/daemons/CMakeLists.txt b/src/daemons/CMakeLists.txt
@@ -146,6 +146,7 @@ nebula_add_executable(
         $<TARGET_OBJECTS:graph_obj>
         $<TARGET_OBJECTS:ft_es_graph_adapter_obj>
         $<TARGET_OBJECTS:storage_thrift_obj>
+        $<TARGET_OBJECTS:gc_obj>
         ${common_deps}
     LIBRARIES
         ${PROXYGEN_LIBRARIES}
@@ -245,6 +246,7 @@ nebula_add_executable(
         $<TARGET_OBJECTS:meta_version_man_obj>
         $<TARGET_OBJECTS:meta_data_upgrade_obj>
         $<TARGET_OBJECTS:meta_v2_thrift_obj>
+        $<TARGET_OBJECTS:gc_obj>
         ${storage_meta_deps}
         ${common_deps}
     LIBRARIES

diff --git a/src/graph/CMakeLists.txt b/src/graph/CMakeLists.txt
@@ -13,3 +13,4 @@ nebula_add_subdirectory(stats)
 nebula_add_subdirectory(util)
 nebula_add_subdirectory(validator)
 nebula_add_subdirectory(visitor)
+nebula_add_subdirectory(gc)
diff --git a/src/graph/context/CMakeLists.txt b/src/graph/context/CMakeLists.txt
@@ -1,7 +1,3 @@
-# Copyright (c) 2020 vesoft inc. All rights reserved.
-#
-# This source code is licensed under Apache 2.0 License.
-
 nebula_add_library(
     graph_context_obj OBJECT
     QueryContext.cpp

diff --git a/src/graph/context/ExecutionContext.cpp b/src/graph/context/ExecutionContext.cpp
@@ -5,6 +5,9 @@
 
 #include "graph/context/ExecutionContext.h"
 
+#include "graph/gc/GC.h"
+#include "graph/service/GraphFlags.h"
+
 namespace nebula {
 namespace graph {
 constexpr int64_t ExecutionContext::kLatestVersion;
@@ -23,7 +26,12 @@ void ExecutionContext::setResult(const std::string& name, Result&& result) {
 }
 
 void ExecutionContext::dropResult(const std::string& name) {
-  valueMap_[name].clear();
+  auto& val = valueMap_[name];
+  if (FLAGS_enable_async_gc) {
+    GC::instance().clear(std::move(val));
+  } else {
+    val.clear();
+  }
 }
 
 size_t ExecutionContext::numVersions(const std::string& name) const {

diff --git a/src/graph/context/Iterator.cpp b/src/graph/context/Iterator.cpp
@@ -602,6 +602,15 @@ void GetNeighborsIter::clearEdges() {
   }
 }
 
+SequentialIter::SequentialIter(const SequentialIter& iter)
+    : Iterator(iter.valuePtr(), Kind::kSequential) {
+  auto valuePtr = iter.valuePtr();
+  auto& ds = valuePtr->mutableDataSet();
+  iter_ = ds.rows.begin();
+  rows_ = &ds.rows;
+  colIndices_ = iter.getColIndices();
+}
+
 SequentialIter::SequentialIter(std::shared_ptr<Value> value, bool checkMemory)
     : Iterator(value, Kind::kSequential, checkMemory) {
   DCHECK(value->isDataSet());
@@ -715,6 +724,11 @@ Value SequentialIter::getEdge() const {
   return getColumn("EDGE");
 }
 
+PropIter::PropIter(const PropIter& iter) : SequentialIter(iter) {
+  dsIndex_ = iter.dsIndex_;
+  kind_ = Kind::kProp;
+}
+
 PropIter::PropIter(std::shared_ptr<Value> value, bool checkMemory)
     : SequentialIter(value, checkMemory) {
   DCHECK(value->isDataSet());

diff --git a/src/graph/context/Iterator.h b/src/graph/context/Iterator.h
@@ -422,10 +422,10 @@ class GetNeighborsIter final : public Iterator {
 class SequentialIter : public Iterator {
  public:
   explicit SequentialIter(std::shared_ptr<Value> value, bool checkMemory = false);
+  explicit SequentialIter(const SequentialIter& iter);
 
   // Union multiple sequential iterators
   explicit SequentialIter(std::vector<std::unique_ptr<Iterator>> inputList);
-
   // Union two sequential iterators.
   SequentialIter(std::unique_ptr<Iterator> left, std::unique_ptr<Iterator> right);
 
@@ -527,6 +527,7 @@ class SequentialIter : public Iterator {
 class PropIter final : public SequentialIter {
  public:
   explicit PropIter(std::shared_ptr<Value> value, bool checkMemory = false);
+  explicit PropIter(const PropIter& iter);
 
   std::unique_ptr<Iterator> copy() const override {
     auto copy = std::make_unique<PropIter>(*this);

diff --git a/src/graph/context/test/CMakeLists.txt b/src/graph/context/test/CMakeLists.txt
@@ -47,6 +47,7 @@ SET(CONTEXT_TEST_LIBS
     $<TARGET_OBJECTS:graph_stats_obj>
     $<TARGET_OBJECTS:meta_client_stats_obj>
     $<TARGET_OBJECTS:storage_client_stats_obj>
+    $<TARGET_OBJECTS:gc_obj>
 )
 
 if(ENABLE_STANDALONE_VERSION)

diff --git a/src/graph/executor/Executor.cpp b/src/graph/executor/Executor.cpp
@@ -7,6 +7,7 @@
 #include <folly/String.h>
 #include <folly/executors/InlineExecutor.h>
 
+#include <algorithm>
 #include <atomic>
 
 #include "common/base/ObjectPool.h"
@@ -665,7 +666,6 @@ void Executor::drop(const PlanNode *node) {
         // Make sure drop happened-after count decrement
         CHECK_EQ(inputVar->userCount.load(std::memory_order_acquire), 0);
         ectx_->dropResult(inputVar->name);
-        VLOG(1) << node->kind() << " Drop variable " << inputVar->name;
       }
     }
   }
@@ -734,5 +734,14 @@ folly::Executor *Executor::runner() const {
   return qctx()->rctx()->runner();
 }
 
+size_t Executor::getBatchSize(size_t totalSize) const {
+  // batch size should be the greater one of FLAGS_min_batch_size and (totalSize/FLAGS_max_job_size)
+  size_t jobSize = FLAGS_max_job_size;
+  size_t minBatchSize = FLAGS_min_batch_size;
+  size_t batchSizeTmp = std::ceil(static_cast<float>(totalSize) / jobSize);
+  size_t batchSize = batchSizeTmp > minBatchSize ? batchSizeTmp : minBatchSize;
+  return batchSize;
+}
+
 }  // namespace graph
 }  // namespace nebula
diff --git a/src/graph/executor/Executor.h b/src/graph/executor/Executor.h
@@ -106,6 +106,17 @@ class Executor : private boost::noncopyable, private cpp::NonMovable {
   // Store the default result which not used for later executor
   Status finish(Value &&value);
 
+  size_t getBatchSize(size_t totalSize) const;
+
+  // ScatterFunc: A callback function that handle partial records of a dataset.
+  // GatherFunc: A callback function that gather all results of ScatterFunc, and do post works.
+  // Iterator: An iterator of a dataset.
+  template <
+      class ScatterFunc,
+      class ScatterResult = typename std::result_of<ScatterFunc(size_t, size_t, Iterator *)>::type,
+      class GatherFunc>
+  auto runMultiJobs(ScatterFunc &&scatter, GatherFunc &&gather, Iterator *iter);
+
   int64_t id_;
 
   // Executor name
@@ -129,6 +140,34 @@ class Executor : private boost::noncopyable, private cpp::NonMovable {
   std::unordered_map<std::string, std::string> otherStats_;
 };
 
+template <class ScatterFunc, class ScatterResult, class GatherFunc>
+auto Executor::runMultiJobs(ScatterFunc &&scatter, GatherFunc &&gather, Iterator *iter) {
+  size_t totalSize = iter->size();
+  size_t batchSize = getBatchSize(totalSize);
+
+  // Start multiple jobs for handling the results
+  std::vector<folly::Future<ScatterResult>> futures;
+  size_t begin = 0, end = 0, dispathedCnt = 0;
+  while (dispathedCnt < totalSize) {
+    end = begin + batchSize > totalSize ? totalSize : begin + batchSize;
+    futures.emplace_back(folly::via(
+        runner(),
+        [begin, end, tmpIter = iter->copy(), f = std::move(scatter)]() mutable -> ScatterResult {
+          // Since not all iterators are linear, so iterates to the begin pos
+          size_t tmp = 0;
+          for (; tmpIter->valid() && tmp < begin; ++tmp) {
+            tmpIter->next();
+          }
+
+          return f(begin, end, tmpIter.get());
+        }));
+    begin = end;
+    dispathedCnt += batchSize;
+  }
+
+  // Gather all results and do post works
+  return folly::collect(futures).via(runner()).thenValue(std::move(gather));
+}
 }  // namespace graph
 }  // namespace nebula
 

diff --git a/src/graph/executor/algo/ShortestPathBase.cpp b/src/graph/executor/algo/ShortestPathBase.cpp
@@ -218,7 +218,9 @@ void ShortestPathBase::addStats(PropRpcResponse& resp, int64_t timeInUSec) const
     ss << "\n}";
   }
   ss << "\n}";
+  statsLock_.lock();
   stats_->emplace(folly::sformat("get_prop "), ss.str());
+  statsLock_.unlock();
 }
 
 }  // namespace graph