chore: lock free info command with replicaof v2

Signed-off-by: kostas <[email protected]>

Author: kostasrim

src/server/cluster/outgoing_slot_migration.cc

@@ -166,6 +166,7 @@ void OutgoingMigration::Finish(const GenericError& error) {
     switch (state_) {
       case MigrationState::C_FATAL:
       case MigrationState::C_FINISHED:
+        CloseSocket();
         return;  // Already finished, nothing else to do
 
       case MigrationState::C_CONNECTING:
@@ -192,6 +193,9 @@ void OutgoingMigration::Finish(const GenericError& error) {
     });
     exec_st_.JoinErrorHandler();
   }
+
+  // Close socket for clean disconnect.
+  CloseSocket();
 }
 
 MigrationState OutgoingMigration::GetState() const {

src/server/protocol_client.cc

@@ -109,11 +109,16 @@ ProtocolClient::ProtocolClient(string host, uint16_t port) {
 #ifdef DFLY_USE_SSL
   MaybeInitSslCtx();
 #endif
+  // We initialize the proactor thread here such that it never races with Sock().
+  // ProtocolClient is never migrated to a different thread, so this is safe.
+  socket_thread_ = ProactorBase::me();
 }
+
 ProtocolClient::ProtocolClient(ServerContext context) : server_context_(std::move(context)) {
 #ifdef DFLY_USE_SSL
   MaybeInitSslCtx();
 #endif
+  socket_thread_ = ProactorBase::me();
 }
 
 ProtocolClient::~ProtocolClient() {
@@ -162,6 +167,7 @@ error_code ProtocolClient::ResolveHostDns() {
 error_code ProtocolClient::ConnectAndAuth(std::chrono::milliseconds connect_timeout_ms,
                                           ExecutionState* cntx) {
   ProactorBase* mythread = ProactorBase::me();
+  DCHECK(mythread == socket_thread_);
   CHECK(mythread);
   {
     unique_lock lk(sock_mu_);
@@ -235,6 +241,9 @@ void ProtocolClient::CloseSocket() {
         auto ec = sock_->Shutdown(SHUT_RDWR);
         LOG_IF(ERROR, ec) << "Could not shutdown socket " << ec;
       }
+      auto ec = sock_->Close();  // Quietly close.
+
+      LOG_IF(WARNING, ec) << "Error closing socket " << ec << "/" << ec.message();
     });
   }
 }

src/server/protocol_client.h

@@ -45,6 +45,13 @@ class ProtocolClient {
   uint64_t LastIoTime() const;
   void TouchIoTime();
 
+  // Used to set the socket_thread_ prior the initialization of sock_.
+  // That way Proactor() returns the right thread even when the sock_ is
+  // not yet initialized
+  void SetSocketThread(util::fb2::ProactorBase* sock_thread) {
+    socket_thread_ = sock_thread;
+  }
+
  protected:
   struct ServerContext {
     std::string host;
@@ -107,7 +114,7 @@ class ProtocolClient {
   }
 
   auto* Proactor() const {
-    return sock_->proactor();
+    return socket_thread_;
   }
 
   util::FiberSocketBase* Sock() const {
@@ -142,6 +149,7 @@ class ProtocolClient {
 #else
   void* ssl_ctx_{nullptr};
 #endif
+  util::fb2::ProactorBase* socket_thread_;
 };
 
 }  // namespace dfly

src/server/replica.cc

@@ -170,9 +170,12 @@ std::optional<Replica::LastMasterSyncData> Replica::Stop() {
   sync_fb_.JoinIfNeeded();
   DVLOG(1) << "MainReplicationFb stopped " << this;
   acks_fb_.JoinIfNeeded();
-  for (auto& flow : shard_flows_) {
-    flow.reset();
-  }
+
+  proactor_->Await([this]() {
+    for (auto& flow : shard_flows_) {
+      flow.reset();
+    }
+  });
 
   if (last_journal_LSNs_.has_value()) {
     return LastMasterSyncData{master_context_.master_repl_id, last_journal_LSNs_.value()};
@@ -501,29 +504,45 @@ error_code Replica::InitiatePSync() {
   return error_code{};
 }
 
-// Initialize and start sub-replica for each flow.
-error_code Replica::InitiateDflySync(std::optional<LastMasterSyncData> last_master_sync_data) {
-  auto start_time = absl::Now();
-
-  // Initialize MultiShardExecution.
-  multi_shard_exe_.reset(new MultiShardExecution());
+void Replica::InitializeShardFlows() {
+  decltype(shard_flows_) shard_flows_copy;
+  shard_flows_copy.resize(master_context_.num_flows);
+  DCHECK(!shard_flows_copy.empty());
+  thread_flow_map_ = Partition(shard_flows_copy.size());
+  const size_t pool_sz = shard_set->pool()->size();
 
-  // Initialize shard flows.
-  shard_flows_.resize(master_context_.num_flows);
-  DCHECK(!shard_flows_.empty());
-  for (unsigned i = 0; i < shard_flows_.size(); ++i) {
-    // Transfer LSN state for partial sync
+  for (size_t i = 0; i < shard_flows_copy.size(); ++i) {
     uint64_t partial_sync_lsn = 0;
-    if (shard_flows_[i]) {
+    if (!shard_flows_.empty() && shard_flows_[i]) {
       partial_sync_lsn = shard_flows_[i]->JournalExecutedCount();
     }
-    shard_flows_[i].reset(
+    shard_flows_copy[i].reset(
         new DflyShardReplica(server(), master_context_, i, &service_, multi_shard_exe_));
     if (partial_sync_lsn > 0) {
       shard_flows_[i]->SetRecordsExecuted(partial_sync_lsn);
     }
   }
-  thread_flow_map_ = Partition(shard_flows_.size());
+
+  shard_set->pool()->AwaitFiberOnAll([pool_sz, this, &shard_flows_copy](auto index, auto* ctx) {
+    for (unsigned i = index; i < shard_flows_copy.size(); i += pool_sz) {
+      shard_flows_copy[i]->SetSocketThread(ProactorBase::me());
+    }
+  });
+  // now update shard_flows on proactor thread
+  shard_flows_ = std::move(shard_flows_copy);
+}
+
+// Initialize and start sub-replica for each flow.
+error_code Replica::InitiateDflySync(std::optional<LastMasterSyncData> last_master_sync_data) {
+  auto start_time = absl::Now();
+
+  // Initialize MultiShardExecution.
+  multi_shard_exe_.reset(new MultiShardExecution());
+
+  // Initialize shard flows. The update to the shard_flows_ should be done by this thread.
+  // Otherwise, there is a race condition between GetSummary() and the shard_flows_[i].reset()
+  // below.
+  InitializeShardFlows();
 
   // Blocked on until all flows got full sync cut.
   BlockingCounter sync_block{unsigned(shard_flows_.size())};
@@ -1215,6 +1234,7 @@ auto Replica::GetSummary() const -> Summary {
     // Note: we access LastIoTime from foreigh thread in unsafe manner. However, specifically here
     // it's unlikely to cause a real bug.
     for (const auto& flow : shard_flows_) {  // Get last io time from all sub flows.
+      DCHECK(Proactor() == ProactorBase::me());
       last_io_time = std::max(last_io_time, flow->LastIoTime());
     }
 
@@ -1246,25 +1266,14 @@ auto Replica::GetSummary() const -> Summary {
     return res;
   };
 
-  if (Sock())
-    return Proactor()->AwaitBrief(f);
-
-  /**
-   * when this branch happens: there is a very short grace period
-   * where Sock() is not initialized, yet the server can
-   * receive ROLE/INFO commands. That period happens when launching
-   * an instance with '--replicaof' and then immediately
-   * sending a command.
-   *
-   * In that instance, we have to run f() on the current fiber.
-   */
-  return f();
+  return Proactor()->AwaitBrief(f);
 }
 
 std::vector<uint64_t> Replica::GetReplicaOffset() const {
   std::vector<uint64_t> flow_rec_count;
   flow_rec_count.resize(shard_flows_.size());
   for (const auto& flow : shard_flows_) {
+    DCHECK(flow.get());
     uint32_t flow_id = flow->FlowId();
     uint64_t rec_count = flow->JournalExecutedCount();
     DCHECK_LT(flow_id, shard_flows_.size());

src/server/replica.h

@@ -154,6 +154,8 @@ class Replica : ProtocolClient {
   size_t GetRecCountExecutedPerShard(const std::vector<unsigned>& indexes) const;
 
  private:
+  void InitializeShardFlows();
+
   util::fb2::ProactorBase* proactor_ = nullptr;
   Service& service_;
   MasterContext master_context_;