From 0e1ba927bdb0c07fbb056979267144339ef4ae95 Mon Sep 17 00:00:00 2001
From: Antonio Andelic <antonio@clickhouse.com>
Date: Fri, 8 Apr 2022 07:18:18 +0000
Subject: [PATCH] Add argument for force recovery

---
 programs/keeper/Keeper.cpp        | 18 ++++++++++++++++++
 programs/keeper/Keeper.h          |  2 ++
 src/Coordination/KeeperServer.cpp | 13 +++++++++----
 3 files changed, 29 insertions(+), 4 deletions(-)

diff --git a/programs/keeper/Keeper.cpp b/programs/keeper/Keeper.cpp
index 1d9bbef58a5..d5af2d6148e 100644
--- a/programs/keeper/Keeper.cpp
+++ b/programs/keeper/Keeper.cpp
@@ -239,6 +239,18 @@ std::string Keeper::getDefaultConfigFileName() const
     return "keeper_config.xml";
 }
 
+void Keeper::handleCustomArguments(const std::string & arg, const std::string & value)
+{
+    if (arg == "force-recovery")
+    {
+        assert(value.empty());
+        config().setBool("keeper_server.recover", true);
+        return;
+    }
+
+    throw Exception(ErrorCodes::LOGICAL_ERROR, "Invalid argument {} provided", arg);
+}
+
 void Keeper::defineOptions(Poco::Util::OptionSet & options)
 {
     options.addOption(
@@ -251,6 +263,12 @@ void Keeper::defineOptions(Poco::Util::OptionSet & options)
             .required(false)
             .repeatable(false)
             .binding("version"));
+    options.addOption(
+        Poco::Util::Option("force-recovery", "force-recovery", "Force recovery mode allowing Keeper to overwrite cluster configuration")
+        .required(false)
+        .repeatable(false)
+        .noArgument()
+        .callback(Poco::Util::OptionCallback<Keeper>(this, &Keeper::handleCustomArguments)));
     BaseDaemon::defineOptions(options);
 }
 
diff --git a/programs/keeper/Keeper.h b/programs/keeper/Keeper.h
index 5b8fbadd0a2..db7a117c2e2 100644
--- a/programs/keeper/Keeper.h
+++ b/programs/keeper/Keeper.h
@@ -43,6 +43,8 @@ public:
 protected:
     void logRevision() const override;
 
+    void handleCustomArguments(const std::string & arg, const std::string & value);
+
     int run() override;
 
     void initialize(Application & self) override;
diff --git a/src/Coordination/KeeperServer.cpp b/src/Coordination/KeeperServer.cpp
index 25d985a7ea3..13d7f67e697 100644
--- a/src/Coordination/KeeperServer.cpp
+++ b/src/Coordination/KeeperServer.cpp
@@ -102,7 +102,7 @@ KeeperServer::KeeperServer(
                         checkAndGetSuperdigest(configuration_and_settings_->super_digest)))
     , state_manager(nuraft::cs_new<KeeperStateManager>(server_id, "keeper_server", configuration_and_settings_->log_storage_path, config, coordination_settings))
     , log(&Poco::Logger::get("KeeperServer"))
-    , recover(config.getBool("recover"))
+    , recover(config.has("keeper_server.recover") && config.getBool("keeper_server.recover"))
 {
     if (coordination_settings->quorum_reads)
         LOG_WARNING(log, "Quorum reads enabled, Keeper will work slower.");
@@ -127,7 +127,6 @@ void KeeperServer::loadLatestConfig()
 
 
         state_manager->save_config(*latest_log_store_config);
-        state_machine->commit_config(latest_log_store_config->get_log_idx(), latest_log_store_config);
         return;
     }
 
@@ -182,6 +181,13 @@ void KeeperServer::startup(bool enable_ipv6)
 
     params.return_method_ = nuraft::raft_params::async_handler;
 
+    if (recover)
+    {
+        LOG_INFO(log, "Custom quorum size");
+        params.with_custom_commit_quorum_size(1);
+        params.with_custom_election_quorum_size(1);
+    }
+
     nuraft::asio_service::options asio_opts{};
     if (state_manager->isSecure())
     {
@@ -226,13 +232,12 @@ void KeeperServer::launchRaftServer(
     nuraft::ptr<nuraft::state_mgr> casted_state_manager = state_manager;
     nuraft::ptr<nuraft::state_machine> casted_state_machine = state_machine;
 
-    loadLatestConfig();
-
     /// raft_server creates unique_ptr from it
     nuraft::context * ctx = new nuraft::context(
         casted_state_manager, casted_state_machine,
         asio_listener, logger, rpc_cli_factory, scheduler, params);
 
+    loadLatestConfig();
     raft_instance = nuraft::cs_new<nuraft::raft_server>(ctx, init_options);
 
     raft_instance->start_server(init_options.skip_initial_election_timeout_);