SERVER-49400 Differenciate VectorClock persist behavior based on cluster role

7 files changed, 148 insertions, 6 deletions

diff --git a/jstests/core/views/views_all_commands.js b/jstests/core/views/views_all_commands.js
index a9952cae3b5..672cdd38f97 100644
--- a/jstests/core/views/views_all_commands.js
+++ b/jstests/core/views/views_all_commands.js
@@ -118,6 +118,7 @@ let viewsCommandTests = {
     _shardsvrRenameCollection: {skip: isAnInternalCommand},
     _shardsvrShardCollection: {skip: isAnInternalCommand},
     _transferMods: {skip: isAnInternalCommand},
+    _vectorClockPersist: {skip: isAnInternalCommand},
     abortTransaction: {skip: isUnrelated},
     addShard: {skip: isUnrelated},
     addShardToZone: {skip: isUnrelated},
diff --git a/jstests/replsets/db_reads_while_recovering_all_commands.js b/jstests/replsets/db_reads_while_recovering_all_commands.js
index 9e094c8ba79..1d7c13742df 100644
--- a/jstests/replsets/db_reads_while_recovering_all_commands.js
+++ b/jstests/replsets/db_reads_while_recovering_all_commands.js
@@ -66,6 +66,7 @@ const allCommands = {
     _shardsvrMovePrimary: {skip: isPrimaryOnly},
     _shardsvrShardCollection: {skip: isPrimaryOnly},
     _transferMods: {skip: isPrimaryOnly},
+    _vectorClockPersist: {skip: isPrimaryOnly},
     abortTransaction: {skip: isPrimaryOnly},
     aggregate: {
         command: {aggregate: collName, pipeline: [{$match: {}}], cursor: {}},
diff --git a/jstests/sharding/read_write_concern_defaults_application.js b/jstests/sharding/read_write_concern_defaults_application.js
index 4768ed04aa7..d95cdd6eec7 100644
--- a/jstests/sharding/read_write_concern_defaults_application.js
+++ b/jstests/sharding/read_write_concern_defaults_application.js
@@ -117,6 +117,7 @@ let testCases = {
     _shardsvrRenameCollection: {skip: "internal command"},
     _shardsvrShardCollection: {skip: "internal command"},
     _transferMods: {skip: "internal command"},
+    _vectorClockPersist: {skip: "internal command"},
     abortTransaction: {
         setUp: function(conn) {
             assert.commandWorked(conn.getDB(db).runCommand({create: coll, writeConcern: {w: 1}}));
diff --git a/src/mongo/db/SConscript b/src/mongo/db/SConscript
index c5a203c7163..991bd28507e 100644
--- a/src/mongo/db/SConscript
+++ b/src/mongo/db/SConscript
@@ -1623,6 +1623,7 @@ env.Library(
 env.Library(
     target='vector_clock_mongod',
     source=[
+        'vector_clock_persist_command.cpp',
         'vector_clock_mongod.cpp',
     ],
     LIBDEPS=[
@@ -1633,6 +1634,8 @@ env.Library(
         'rw_concern_d',
         'repl/replica_set_aware_service',
         'server_options_core',
+        '$BUILD_DIR/mongo/db/s/sharding_api_d',
+        '$BUILD_DIR/mongo/s/grid',
     ],
 )
 
diff --git a/src/mongo/db/vector_clock_mongod.cpp b/src/mongo/db/vector_clock_mongod.cpp
index 51a826a71ee..1b34ca5133d 100644
--- a/src/mongo/db/vector_clock_mongod.cpp
+++ b/src/mongo/db/vector_clock_mongod.cpp
@@ -33,10 +33,13 @@
 #include "mongo/db/logical_time_validator.h"
 #include "mongo/db/persistent_task_store.h"
 #include "mongo/db/repl/replica_set_aware_service.h"
+#include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/s/sharding_state.h"
 #include "mongo/db/vector_clock_document_gen.h"
 #include "mongo/db/vector_clock_mutable.h"
 #include "mongo/executor/task_executor_pool.h"
 #include "mongo/logv2/log.h"
+#include "mongo/s/grid.h"
 #include "mongo/util/concurrency/thread_pool.h"
 
 namespace mongo {
@@ -247,7 +250,34 @@ private:
         }
 
         std::string getOperationName() override {
-            return "persist";
+            return "localPersist";
+        }
+    };
+
+    /*
+     * VectorClockStateOperation invoking PersistOperation on a shard server's primary.
+     */
+    class RemotePersistOperation : public VectorClockStateOperation {
+        void execute(VectorClockMongoD* vectorClock, OperationContext* opCtx) override {
+            auto const shardingState = ShardingState::get(opCtx);
+            invariant(shardingState->enabled());
+
+            auto selfShard = uassertStatusOK(
+                Grid::get(opCtx)->shardRegistry()->getShard(opCtx, shardingState->shardId()));
+
+            auto cmdResponse = uassertStatusOK(selfShard->runCommandWithFixedRetryAttempts(
+                opCtx,
+                ReadPreferenceSetting{ReadPreference::PrimaryOnly},
+                NamespaceString::kVectorClockNamespace.toString(),
+                BSON("_vectorClockPersist" << 1),
+                Seconds{30},
+                Shard::RetryPolicy::kIdempotent));
+
+            uassertStatusOK(cmdResponse.commandStatus);
+        }
+
+        std::string getOperationName() override {
+            return "remotePersist";
         }
     };
 
@@ -285,6 +315,7 @@ private:
     };
 
     PersistOperation _persistOperation;
+    RemotePersistOperation _remotePersistOperation;
     RecoverOperation _recoverOperation;
 };
 
@@ -431,7 +462,17 @@ void VectorClockMongoD::_recoverComponent(OperationContext* opCtx,
 }
 
 SharedSemiFuture<void> VectorClockMongoD::persist(OperationContext* opCtx) {
-    return _persistOperation.performOperation(this, opCtx->getServiceContext());
+    if (serverGlobalParams.clusterRole == ClusterRole::ShardServer) {
+        const auto replCoord = repl::ReplicationCoordinator::get(opCtx);
+
+        if (replCoord->getMemberState().primary()) {
+            return _persistOperation.performOperation(this, opCtx->getServiceContext());
+        }
+
+        return _remotePersistOperation.performOperation(this, opCtx->getServiceContext());
+    }
+
+    return SharedSemiFuture<void>();
 }
 
 void VectorClockMongoD::waitForInMemoryVectorClockToBePersisted(OperationContext* opCtx) {
diff --git a/src/mongo/db/vector_clock_mongod_test.cpp b/src/mongo/db/vector_clock_mongod_test.cpp
index 6b8a9d663f0..d59d947bd90 100644
--- a/src/mongo/db/vector_clock_mongod_test.cpp
+++ b/src/mongo/db/vector_clock_mongod_test.cpp
@@ -33,6 +33,7 @@
 #include "mongo/db/keys_collection_manager.h"
 #include "mongo/db/logical_time_validator.h"
 #include "mongo/db/persistent_task_store.h"
+#include "mongo/db/repl/replication_coordinator_mock.h"
 #include "mongo/db/s/sharding_mongod_test_fixture.h"
 #include "mongo/db/vector_clock_document_gen.h"
 #include "mongo/db/vector_clock_mutable.h"
@@ -84,6 +85,22 @@ private:
     std::shared_ptr<KeysCollectionManager> _keyManager;
 };
 
+class VectorClockMongoDTestPrimary : public VectorClockMongoDTest {
+protected:
+    void setUp() override {
+        VectorClockMongoDTest::setUp();
+
+        serverGlobalParams.clusterRole = ClusterRole::ShardServer;
+
+        auto replCoord = repl::ReplicationCoordinator::get(operationContext());
+        ASSERT_OK(replCoord->setFollowerMode(repl::MemberState::RS_PRIMARY));
+    }
+
+    void tearDown() override {
+        VectorClockMongoDTest::tearDown();
+    }
+};
+
 TEST_F(VectorClockMongoDTest, TickClusterTime) {
     auto sc = getServiceContext();
     auto vc = VectorClockMutable::get(sc);
@@ -281,7 +298,7 @@ TEST_F(VectorClockMongoDTest, GossipInExternal) {
     ASSERT_EQ(afterTime3.topologyTime().asTimestamp(), Timestamp(0, 0));
 }
 
-TEST_F(VectorClockMongoDTest, PersistVectorClockDocument) {
+TEST_F(VectorClockMongoDTestPrimary, PersistVectorClockDocument) {
     auto sc = getServiceContext();
     auto opCtx = operationContext();
 
@@ -306,7 +323,7 @@ TEST_F(VectorClockMongoDTest, PersistVectorClockDocument) {
     ASSERT_EQUALS(store.count(opCtx, VectorClock::stateQuery()), 1);
 }
 
-TEST_F(VectorClockMongoDTest, RecoverVectorClockDocument) {
+TEST_F(VectorClockMongoDTestPrimary, RecoverVectorClockDocument) {
     auto sc = getServiceContext();
     auto opCtx = operationContext();
     const auto configTime = LogicalTime(Timestamp(3, 3));
@@ -335,7 +352,7 @@ TEST_F(VectorClockMongoDTest, RecoverVectorClockDocument) {
     ASSERT_EQUALS(actualTopologyTime, topologyTime);
 }
 
-TEST_F(VectorClockMongoDTest, RecoverNotExistingVectorClockDocument) {
+TEST_F(VectorClockMongoDTestPrimary, RecoverNotExistingVectorClockDocument) {
     auto sc = getServiceContext();
     auto opCtx = operationContext();
     auto vc = VectorClockMutable::get(sc);
@@ -364,7 +381,7 @@ TEST_F(VectorClockMongoDTest, RecoverNotExistingVectorClockDocument) {
     ASSERT_EQUALS(actualTopologyTime, topologyTime);
 }
 
-TEST_F(VectorClockMongoDTest, SubsequentPersistRecoverVectorClockDocument) {
+TEST_F(VectorClockMongoDTestPrimary, SubsequentPersistRecoverVectorClockDocument) {
     auto sc = getServiceContext();
     auto opCtx = operationContext();
     auto vc = VectorClockMutable::get(sc);
diff --git a/src/mongo/db/vector_clock_persist_command.cpp b/src/mongo/db/vector_clock_persist_command.cpp
new file mode 100644
index 00000000000..6635898febf
--- /dev/null
+++ b/src/mongo/db/vector_clock_persist_command.cpp
@@ -0,0 +1,78 @@
+/**
+ *    Copyright (C) 2020-present MongoDB, Inc.
+ *
+ *    This program is free software: you can redistribute it and/or modify
+ *    it under the terms of the Server Side Public License, version 1,
+ *    as published by MongoDB, Inc.
+ *
+ *    This program is distributed in the hope that it will be useful,
+ *    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *    Server Side Public License for more details.
+ *
+ *    You should have received a copy of the Server Side Public License
+ *    along with this program. If not, see
+ *    <http://www.mongodb.com/licensing/server-side-public-license>.
+ *
+ *    As a special exception, the copyright holders give permission to link the
+ *    code of portions of this program with the OpenSSL library under certain
+ *    conditions as described in each individual source file and distribute
+ *    linked combinations including the program with the OpenSSL library. You
+ *    must comply with the Server Side Public License in all respects for
+ *    all of the code used other than as permitted herein. If you modify file(s)
+ *    with this exception, you may extend this exception to your version of the
+ *    file(s), but you are not obligated to do so. If you do not wish to do so,
+ *    delete this exception statement from your version. If you delete this
+ *    exception statement from all source files in the program, then also delete
+ *    it in the license file.
+ */
+
+#define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kCommand
+
+#include "mongo/platform/basic.h"
+
+#include <set>
+
+#include "mongo/db/commands.h"
+#include "mongo/db/repl/replication_coordinator.h"
+#include "mongo/db/vector_clock.h"
+
+namespace mongo {
+namespace {
+
+/**
+ * Internal sharding command run on shard primaries to persist the vector clock state.
+ */
+class VectorClockPersistCommand : public BasicCommand {
+public:
+    VectorClockPersistCommand() : BasicCommand("_vectorClockPersist") {}
+
+    AllowedOnSecondary secondaryAllowed(ServiceContext*) const override {
+        return AllowedOnSecondary::kNever;
+    }
+
+    virtual bool supportsWriteConcern(const BSONObj& cmd) const override {
+        return true;
+    }
+
+    std::string help() const override {
+        return "Internal sharding command run on shard primaries to persist the vector clock "
+               "state.";
+    }
+
+    bool run(OperationContext* opCtx,
+             const std::string& dbname_unused,
+             const BSONObj& cmdObj,
+             BSONObjBuilder& result) {
+        auto sc = opCtx->getServiceContext();
+        auto vc = VectorClock::get(sc);
+
+        vc->persist(opCtx).get(opCtx);
+
+        return true;
+    }
+
+} vectorClockPersistCmd;
+
+}  // namespace
+}  // namespace mongo