Skip to main content

cmd/
standalone.rs

1// Copyright 2023 Greptime Team
2//
3// Licensed under the Apache License, Version 2.0 (the "License");
4// you may not use this file except in compliance with the License.
5// You may obtain a copy of the License at
6//
7//     http://www.apache.org/licenses/LICENSE-2.0
8//
9// Unless required by applicable law or agreed to in writing, software
10// distributed under the License is distributed on an "AS IS" BASIS,
11// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12// See the License for the specific language governing permissions and
13// limitations under the License.
14
15use std::fmt::Debug;
16use std::net::SocketAddr;
17use std::path::Path;
18use std::sync::Arc;
19use std::{fs, path};
20
21use async_trait::async_trait;
22use cache::{build_fundamental_cache_registry, with_default_composite_cache_registry};
23use catalog::information_schema::InformationExtensionRef;
24use catalog::kvbackend::{CatalogManagerConfiguratorRef, KvBackendCatalogManagerBuilder};
25use catalog::process_manager::ProcessManager;
26use clap::Parser;
27use common_base::Plugins;
28use common_catalog::consts::{MIN_USER_FLOW_ID, MIN_USER_TABLE_ID};
29use common_config::{Configurable, metadata_store_dir};
30use common_error::ext::BoxedError;
31use common_meta::cache::LayeredCacheRegistryBuilder;
32use common_meta::ddl::flow_meta::FlowMetadataAllocator;
33use common_meta::ddl::table_meta::TableMetadataAllocator;
34use common_meta::ddl::{DdlContext, NoopRegionFailureDetectorControl};
35use common_meta::ddl_manager::{DdlManager, DdlManagerConfiguratorRef, DdlManagerRef};
36use common_meta::key::flow::FlowMetadataManager;
37use common_meta::key::{TableMetadataManager, TableMetadataManagerRef};
38use common_meta::kv_backend::KvBackendRef;
39use common_meta::node_manager::{FlownodeRef, NodeManagerRef};
40use common_meta::procedure_executor::{LocalProcedureExecutor, ProcedureExecutorRef};
41use common_meta::region_keeper::MemoryRegionKeeper;
42use common_meta::region_registry::LeaderRegionRegistry;
43use common_meta::sequence::{Sequence, SequenceBuilder};
44use common_meta::wal_provider::{WalProviderRef, build_wal_provider};
45use common_options::plugin_options::StandaloneFlag;
46use common_procedure::ProcedureManagerRef;
47use common_query::prelude::set_default_prefix;
48use common_telemetry::info;
49use common_telemetry::logging::{DEFAULT_LOGGING_DIR, TracingOptions};
50use common_time::timezone::set_default_timezone;
51use common_version::{short_version, verbose_version};
52use datanode::config::DatanodeOptions;
53use datanode::datanode::{Datanode, DatanodeBuilder};
54use datanode::region_server::RegionServer;
55use flow::{
56    FlownodeBuilder, FlownodeInstance, FlownodeOptions, FrontendClient, FrontendInvoker,
57    GrpcQueryHandlerWithBoxedError,
58};
59use frontend::frontend::Frontend;
60use frontend::instance::StandaloneDatanodeManager;
61use frontend::instance::builder::FrontendBuilder;
62use frontend::server::Services;
63use meta_srv::metasrv::{FLOW_ID_SEQ, TABLE_ID_SEQ};
64use plugins::PluginOptions;
65use plugins::frontend::context::{
66    CatalogManagerConfigureContext, StandaloneCatalogManagerConfigureContext,
67};
68use plugins::standalone::context::DdlManagerConfigureContext;
69use servers::tls::{TlsMode, TlsOption, merge_tls_option};
70use snafu::ResultExt;
71use standalone::options::StandaloneOptions;
72use standalone::{StandaloneInformationExtension, StandaloneRepartitionProcedureFactory};
73use tracing_appender::non_blocking::WorkerGuard;
74
75use crate::error::{OtherSnafu, Result, StartFlownodeSnafu};
76use crate::options::{GlobalOptions, GreptimeOptions};
77use crate::{App, create_resource_limit_metrics, error, log_versions, maybe_activate_heap_profile};
78
79pub const APP_NAME: &str = "greptime-standalone";
80
81#[derive(Parser)]
82pub struct Command {
83    #[clap(subcommand)]
84    subcmd: SubCommand,
85}
86
87impl Command {
88    pub async fn build(&self, opts: GreptimeOptions<StandaloneOptions>) -> Result<Instance> {
89        self.subcmd.build(opts).await
90    }
91
92    pub fn load_options(
93        &self,
94        global_options: &GlobalOptions,
95    ) -> Result<GreptimeOptions<StandaloneOptions>> {
96        self.subcmd.load_options(global_options)
97    }
98}
99
100#[derive(Parser)]
101enum SubCommand {
102    Start(StartCommand),
103}
104
105impl SubCommand {
106    async fn build(&self, opts: GreptimeOptions<StandaloneOptions>) -> Result<Instance> {
107        match self {
108            SubCommand::Start(cmd) => cmd.build(opts).await,
109        }
110    }
111
112    fn load_options(
113        &self,
114        global_options: &GlobalOptions,
115    ) -> Result<GreptimeOptions<StandaloneOptions>> {
116        match self {
117            SubCommand::Start(cmd) => cmd.load_options(global_options),
118        }
119    }
120}
121
122pub struct Instance {
123    datanode: Datanode,
124    frontend: Frontend,
125    flownode: FlownodeInstance,
126    procedure_manager: ProcedureManagerRef,
127    wal_provider: WalProviderRef,
128    leader_services_controller: Box<dyn StandaloneLeaderServicesController>,
129    // Keep the logging guard to prevent the worker from being dropped.
130    _guard: Vec<WorkerGuard>,
131}
132
133impl Instance {
134    /// Find the socket addr of a server by its `name`.
135    pub fn server_addr(&self, name: &str) -> Option<SocketAddr> {
136        self.frontend.server_handlers().addr(name)
137    }
138
139    /// Get the mutable Frontend component of this Standalone instance for externally modification
140    /// by others (might not be in this code base, so don't delete this function).
141    pub fn mut_frontend(&mut self) -> &mut Frontend {
142        &mut self.frontend
143    }
144
145    /// Get the Datanode component of this Standalone instance for externally usage
146    /// by others (might not be in this code base, so don't delete this function).
147    pub fn datanode(&self) -> &Datanode {
148        &self.datanode
149    }
150}
151
152#[async_trait]
153impl App for Instance {
154    fn name(&self) -> &str {
155        APP_NAME
156    }
157
158    async fn start(&mut self) -> Result<()> {
159        self.datanode.start_telemetry();
160
161        self.leader_services_controller
162            .start(
163                self.procedure_manager.clone(),
164                self.wal_provider.clone(),
165                self.datanode.region_server(),
166            )
167            .await?;
168
169        plugins::start_frontend_plugins(self.frontend.instance.plugins().clone())
170            .await
171            .context(error::StartFrontendSnafu)?;
172
173        self.frontend
174            .start()
175            .await
176            .context(error::StartFrontendSnafu)?;
177
178        self.flownode.start().await.context(StartFlownodeSnafu)?;
179
180        Ok(())
181    }
182
183    async fn stop(&mut self) -> Result<()> {
184        self.frontend
185            .shutdown()
186            .await
187            .context(error::ShutdownFrontendSnafu)?;
188
189        self.leader_services_controller
190            .stop(
191                self.procedure_manager.clone(),
192                self.datanode.region_server(),
193            )
194            .await?;
195
196        self.datanode
197            .shutdown()
198            .await
199            .context(error::ShutdownDatanodeSnafu)?;
200
201        self.flownode
202            .shutdown()
203            .await
204            .context(error::ShutdownFlownodeSnafu)?;
205
206        info!("Datanode instance stopped.");
207
208        Ok(())
209    }
210}
211
212#[derive(Debug, Default, Parser)]
213pub struct StartCommand {
214    #[clap(long)]
215    http_addr: Option<String>,
216    #[clap(long = "grpc-bind-addr", alias = "rpc-bind-addr", alias = "rpc-addr")]
217    grpc_bind_addr: Option<String>,
218    #[clap(long)]
219    mysql_addr: Option<String>,
220    #[clap(long)]
221    postgres_addr: Option<String>,
222    #[clap(short, long)]
223    influxdb_enable: bool,
224    #[clap(short, long)]
225    pub config_file: Option<String>,
226    #[clap(long)]
227    tls_mode: Option<TlsMode>,
228    #[clap(long)]
229    tls_cert_path: Option<String>,
230    #[clap(long)]
231    tls_key_path: Option<String>,
232    #[clap(long)]
233    tls_watch: bool,
234    #[clap(long)]
235    user_provider: Option<String>,
236    #[clap(long, default_value = "GREPTIMEDB_STANDALONE")]
237    pub env_prefix: String,
238    /// The working home directory of this standalone instance.
239    #[clap(long)]
240    data_home: Option<String>,
241}
242
243impl StartCommand {
244    /// Load the GreptimeDB options from various sources (command line, config file or env).
245    pub fn load_options(
246        &self,
247        global_options: &GlobalOptions,
248    ) -> Result<GreptimeOptions<StandaloneOptions>> {
249        let mut opts = GreptimeOptions::<StandaloneOptions>::load_layered_options(
250            self.config_file.as_deref(),
251            self.env_prefix.as_ref(),
252        )
253        .context(error::LoadLayeredConfigSnafu)?;
254
255        self.merge_with_cli_options(global_options, &mut opts.component)?;
256        opts.component.sanitize();
257
258        Ok(opts)
259    }
260
261    // The precedence order is: cli > config file > environment variables > default values.
262    pub fn merge_with_cli_options(
263        &self,
264        global_options: &GlobalOptions,
265        opts: &mut StandaloneOptions,
266    ) -> Result<()> {
267        if let Some(dir) = &global_options.log_dir {
268            opts.logging.dir.clone_from(dir);
269        }
270
271        if global_options.log_level.is_some() {
272            opts.logging.level.clone_from(&global_options.log_level);
273        }
274
275        opts.tracing = TracingOptions {
276            #[cfg(feature = "tokio-console")]
277            tokio_console_addr: global_options.tokio_console_addr.clone(),
278        };
279
280        let tls_opts = TlsOption::new(
281            self.tls_mode,
282            self.tls_cert_path.clone(),
283            self.tls_key_path.clone(),
284            self.tls_watch,
285        );
286
287        if let Some(addr) = &self.http_addr {
288            opts.http.addr.clone_from(addr);
289        }
290
291        if let Some(data_home) = &self.data_home {
292            opts.storage.data_home.clone_from(data_home);
293        }
294
295        // If the logging dir is not set, use the default logs dir in the data home.
296        if opts.logging.dir.is_empty() {
297            opts.logging.dir = Path::new(&opts.storage.data_home)
298                .join(DEFAULT_LOGGING_DIR)
299                .to_string_lossy()
300                .to_string();
301        }
302
303        if let Some(addr) = &self.grpc_bind_addr {
304            // frontend grpc addr conflict with datanode default grpc addr
305            let datanode_grpc_addr = DatanodeOptions::default().grpc.bind_addr;
306            if addr.eq(&datanode_grpc_addr) {
307                return error::IllegalConfigSnafu {
308                    msg: format!(
309                        "gRPC listen address conflicts with datanode reserved gRPC addr: {datanode_grpc_addr}",
310                    ),
311                }.fail();
312            }
313            opts.grpc.bind_addr.clone_from(addr);
314            opts.grpc.tls = merge_tls_option(&opts.grpc.tls, tls_opts.clone());
315        }
316
317        if let Some(addr) = &self.mysql_addr {
318            opts.mysql.enable = true;
319            opts.mysql.addr.clone_from(addr);
320            opts.mysql.tls = merge_tls_option(&opts.mysql.tls, tls_opts.clone());
321        }
322
323        if let Some(addr) = &self.postgres_addr {
324            opts.postgres.enable = true;
325            opts.postgres.addr.clone_from(addr);
326            opts.postgres.tls = merge_tls_option(&opts.postgres.tls, tls_opts.clone());
327        }
328
329        if self.influxdb_enable {
330            opts.influxdb.enable = self.influxdb_enable;
331        }
332
333        if let Some(user_provider) = &self.user_provider {
334            opts.user_provider = Some(user_provider.clone());
335        }
336
337        Ok(())
338    }
339
340    #[allow(unreachable_code)]
341    #[allow(unused_variables)]
342    #[allow(clippy::diverging_sub_expression)]
343    /// Build GreptimeDB instance with the loaded options.
344    pub async fn build(&self, opts: GreptimeOptions<StandaloneOptions>) -> Result<Instance> {
345        common_runtime::init_global_runtimes(&opts.runtime);
346
347        let guard = common_telemetry::init_global_logging(
348            APP_NAME,
349            &opts.component.logging,
350            &opts.component.tracing,
351            None,
352            Some(&opts.component.slow_query),
353        );
354
355        log_versions(verbose_version(), short_version(), APP_NAME);
356        maybe_activate_heap_profile(&opts.component.memory);
357        create_resource_limit_metrics(APP_NAME);
358
359        info!("Standalone start command: {:#?}", self);
360        info!("Standalone options: {opts:#?}");
361
362        let (mut instance, _) =
363            Self::build_with(opts.component, opts.plugins, InstanceCreator::default()).await?;
364        instance._guard.extend(guard);
365        Ok(instance)
366    }
367
368    pub async fn build_with(
369        mut opts: StandaloneOptions,
370        plugin_opts: Vec<PluginOptions>,
371        creator: InstanceCreator,
372    ) -> Result<(Instance, InstanceCreatorResult)> {
373        let mut plugins = Plugins::new();
374        plugins.insert(StandaloneFlag);
375        set_default_prefix(opts.default_column_prefix.as_deref())
376            .map_err(BoxedError::new)
377            .context(error::BuildCliSnafu)?;
378
379        opts.grpc.detect_server_addr();
380        let fe_opts = opts.frontend_options();
381        let dn_opts = opts.datanode_options();
382
383        plugins::setup_frontend_plugins(&mut plugins, &plugin_opts, &fe_opts)
384            .await
385            .context(error::StartFrontendSnafu)?;
386
387        plugins::setup_datanode_plugins(&mut plugins, &plugin_opts, &dn_opts)
388            .await
389            .context(error::StartDatanodeSnafu)?;
390
391        set_default_timezone(fe_opts.default_timezone.as_deref())
392            .context(error::InitTimezoneSnafu)?;
393
394        let data_home = &dn_opts.storage.data_home;
395        // Ensure the data_home directory exists.
396        fs::create_dir_all(path::Path::new(data_home))
397            .context(error::CreateDirSnafu { dir: data_home })?;
398
399        let metadata_dir = metadata_store_dir(data_home);
400        let kv_backend = creator
401            .metadata_kv_backend_creator
402            .create(metadata_dir, &opts)
403            .await?;
404        let procedure_manager =
405            standalone::build_procedure_manager(kv_backend.clone(), opts.procedure);
406
407        plugins::setup_standalone_plugins(&mut plugins, &plugin_opts, &opts, kv_backend.clone())
408            .await
409            .context(error::SetupStandalonePluginsSnafu)?;
410
411        // Builds cache registry
412        let layered_cache_builder = LayeredCacheRegistryBuilder::default();
413        let fundamental_cache_registry = build_fundamental_cache_registry(kv_backend.clone());
414        let layered_cache_registry = Arc::new(
415            with_default_composite_cache_registry(
416                layered_cache_builder.add_cache_registry(fundamental_cache_registry),
417            )
418            .context(error::BuildCacheRegistrySnafu)?
419            .build(),
420        );
421
422        let mut builder = DatanodeBuilder::new(dn_opts, plugins.clone(), kv_backend.clone());
423        builder.with_cache_registry(layered_cache_registry.clone());
424        if let Some(writable) = creator.open_regions_writable_override {
425            builder.with_open_regions_writable_override(writable);
426        }
427        let datanode = builder.build().await.context(error::StartDatanodeSnafu)?;
428
429        let information_extension = Arc::new(StandaloneInformationExtension::new(
430            datanode.region_server(),
431            procedure_manager.clone(),
432        ));
433
434        plugins.insert::<InformationExtensionRef>(information_extension.clone());
435
436        let process_manager = Arc::new(ProcessManager::new(opts.grpc.server_addr.clone(), None));
437
438        // for standalone not use grpc, but get a handler to frontend grpc client without
439        // actually make a connection
440        let (frontend_client, frontend_instance_handler) =
441            FrontendClient::from_empty_grpc_handler(opts.query.clone());
442        let frontend_client = Arc::new(frontend_client);
443
444        let builder = KvBackendCatalogManagerBuilder::new(
445            information_extension.clone(),
446            kv_backend.clone(),
447            layered_cache_registry.clone(),
448        )
449        .with_procedure_manager(procedure_manager.clone())
450        .with_process_manager(process_manager.clone());
451        let builder = if let Some(configurator) =
452            plugins.get::<CatalogManagerConfiguratorRef<CatalogManagerConfigureContext>>()
453        {
454            let ctx = StandaloneCatalogManagerConfigureContext {
455                fe_client: frontend_client.clone(),
456            };
457            let ctx = CatalogManagerConfigureContext::Standalone(ctx);
458            configurator
459                .configure(builder, ctx)
460                .await
461                .context(OtherSnafu)?
462        } else {
463            builder
464        };
465        let catalog_manager = builder.build();
466
467        let table_metadata_manager =
468            Self::create_table_metadata_manager(kv_backend.clone()).await?;
469
470        let flow_metadata_manager = Arc::new(FlowMetadataManager::new(kv_backend.clone()));
471        let flownode_options = FlownodeOptions {
472            flow: opts.flow.clone(),
473            ..Default::default()
474        };
475
476        let flow_builder = FlownodeBuilder::new(
477            flownode_options,
478            plugins.clone(),
479            table_metadata_manager.clone(),
480            catalog_manager.clone(),
481            flow_metadata_manager.clone(),
482            frontend_client.clone(),
483        );
484        let flownode = flow_builder
485            .build()
486            .await
487            .map_err(BoxedError::new)
488            .context(error::OtherSnafu)?;
489
490        // set the ref to query for the local flow state
491        {
492            information_extension
493                .set_flow_engine(flownode.flow_engine())
494                .await;
495        }
496
497        let node_manager = creator
498            .node_manager_creator
499            .create(
500                &kv_backend,
501                datanode.region_server(),
502                flownode.flow_engine(),
503            )
504            .await?;
505
506        let table_id_allocator = creator.table_id_allocator_creator.create(&kv_backend);
507        let flow_id_sequence = Arc::new(
508            SequenceBuilder::new(FLOW_ID_SEQ, kv_backend.clone())
509                .initial(MIN_USER_FLOW_ID as u64)
510                .step(10)
511                .build(),
512        );
513        let kafka_options = opts
514            .wal
515            .clone()
516            .try_into()
517            .context(error::InvalidWalProviderSnafu)?;
518        let wal_provider = build_wal_provider(&kafka_options, kv_backend.clone())
519            .await
520            .context(error::BuildWalProviderSnafu)?;
521        let wal_provider = Arc::new(wal_provider);
522        let table_metadata_allocator = Arc::new(TableMetadataAllocator::new(
523            table_id_allocator.clone(),
524            wal_provider.clone(),
525        ));
526        let flow_metadata_allocator = Arc::new(FlowMetadataAllocator::with_noop_peer_allocator(
527            flow_id_sequence,
528        ));
529
530        let ddl_context = DdlContext {
531            node_manager: node_manager.clone(),
532            cache_invalidator: layered_cache_registry.clone(),
533            memory_region_keeper: Arc::new(MemoryRegionKeeper::default()),
534            leader_region_registry: Arc::new(LeaderRegionRegistry::default()),
535            table_metadata_manager: table_metadata_manager.clone(),
536            table_metadata_allocator: table_metadata_allocator.clone(),
537            flow_metadata_manager: flow_metadata_manager.clone(),
538            flow_metadata_allocator: flow_metadata_allocator.clone(),
539            region_failure_detector_controller: Arc::new(NoopRegionFailureDetectorControl),
540        };
541
542        let ddl_manager = DdlManager::try_new(
543            ddl_context,
544            procedure_manager.clone(),
545            Arc::new(StandaloneRepartitionProcedureFactory),
546            true,
547        )
548        .context(error::InitDdlManagerSnafu)?;
549
550        let ddl_manager = if let Some(configurator) =
551            plugins.get::<DdlManagerConfiguratorRef<DdlManagerConfigureContext>>()
552        {
553            let ctx = DdlManagerConfigureContext {
554                kv_backend: kv_backend.clone(),
555                fe_client: frontend_client.clone(),
556                catalog_manager: catalog_manager.clone(),
557            };
558            configurator
559                .configure(ddl_manager, ctx)
560                .await
561                .context(OtherSnafu)?
562        } else {
563            ddl_manager
564        };
565
566        let procedure_executor = creator
567            .procedure_executor_creator
568            .create(Arc::new(ddl_manager), procedure_manager.clone())
569            .await?;
570
571        let fe_instance = FrontendBuilder::new(
572            fe_opts.clone(),
573            kv_backend.clone(),
574            layered_cache_registry.clone(),
575            catalog_manager.clone(),
576            node_manager.clone(),
577            procedure_executor.clone(),
578            process_manager,
579        )
580        .with_plugin(plugins.clone())
581        .try_build()
582        .await
583        .context(error::StartFrontendSnafu)?;
584        let fe_instance = Arc::new(fe_instance);
585
586        // set the frontend client for flownode
587        let grpc_handler = fe_instance.clone() as Arc<dyn GrpcQueryHandlerWithBoxedError>;
588        let weak_grpc_handler = Arc::downgrade(&grpc_handler);
589        frontend_instance_handler
590            .set_handler(weak_grpc_handler)
591            .await;
592
593        // set the frontend invoker for flownode
594        let flow_streaming_engine = flownode.flow_engine().streaming_engine();
595        // flow server need to be able to use frontend to write insert requests back
596        let invoker = FrontendInvoker::build_from(
597            flow_streaming_engine.clone(),
598            catalog_manager.clone(),
599            kv_backend.clone(),
600            layered_cache_registry.clone(),
601            procedure_executor,
602            node_manager.clone(),
603            fe_instance.frontend_peer_addr().to_string(),
604        )
605        .await
606        .context(StartFlownodeSnafu)?;
607        flow_streaming_engine.set_frontend_invoker(invoker).await;
608
609        let servers = Services::new(opts, fe_instance.clone(), plugins.clone())
610            .build()
611            .context(error::StartFrontendSnafu)?;
612
613        let frontend = Frontend {
614            instance: fe_instance,
615            servers,
616            heartbeat_task: None,
617        };
618
619        let instance = Instance {
620            datanode,
621            frontend,
622            flownode,
623            procedure_manager,
624            wal_provider,
625            leader_services_controller: creator.leader_services_controller,
626            _guard: vec![],
627        };
628        let result = InstanceCreatorResult {
629            kv_backend,
630            node_manager,
631            table_id_allocator,
632        };
633        Ok((instance, result))
634    }
635
636    pub async fn create_table_metadata_manager(
637        kv_backend: KvBackendRef,
638    ) -> Result<TableMetadataManagerRef> {
639        let table_metadata_manager = Arc::new(TableMetadataManager::new(kv_backend));
640
641        table_metadata_manager
642            .init()
643            .await
644            .context(error::InitMetadataSnafu)?;
645
646        Ok(table_metadata_manager)
647    }
648}
649
650#[async_trait]
651pub trait NodeManagerCreator: Send + Sync {
652    async fn create(
653        &self,
654        kv_backend: &KvBackendRef,
655        region_server: RegionServer,
656        flow_server: FlownodeRef,
657    ) -> Result<NodeManagerRef>;
658}
659
660pub struct DefaultNodeManagerCreator;
661
662#[async_trait]
663impl NodeManagerCreator for DefaultNodeManagerCreator {
664    async fn create(
665        &self,
666        _: &KvBackendRef,
667        region_server: RegionServer,
668        flow_server: FlownodeRef,
669    ) -> Result<NodeManagerRef> {
670        Ok(Arc::new(StandaloneDatanodeManager {
671            region_server,
672            flow_server,
673        }))
674    }
675}
676
677/// Customizes how standalone opens its metadata KV backend.
678///
679/// The default implementation preserves the built-in raft-engine path. Other
680/// callers can provide a custom implementation without changing standalone
681/// configuration types.
682#[async_trait]
683pub trait MetadataKvBackendCreator: Send + Sync {
684    async fn create(&self, metadata_dir: String, opts: &StandaloneOptions) -> Result<KvBackendRef>;
685}
686
687pub struct DefaultMetadataKvBackendCreator;
688
689#[async_trait]
690impl MetadataKvBackendCreator for DefaultMetadataKvBackendCreator {
691    async fn create(&self, metadata_dir: String, opts: &StandaloneOptions) -> Result<KvBackendRef> {
692        standalone::build_metadata_kvbackend(metadata_dir, opts.metadata_store)
693            .context(error::BuildMetadataKvbackendSnafu)
694    }
695}
696
697pub trait TableIdAllocatorCreator: Send + Sync {
698    fn create(&self, kv_backend: &KvBackendRef) -> Arc<Sequence>;
699}
700
701struct DefaultTableIdAllocatorCreator;
702
703impl TableIdAllocatorCreator for DefaultTableIdAllocatorCreator {
704    fn create(&self, kv_backend: &KvBackendRef) -> Arc<Sequence> {
705        Arc::new(
706            SequenceBuilder::new(TABLE_ID_SEQ, kv_backend.clone())
707                .initial(MIN_USER_TABLE_ID as u64)
708                .step(10)
709                .build(),
710        )
711    }
712}
713
714#[async_trait]
715pub trait ProcedureExecutorCreator: Send + Sync {
716    async fn create(
717        &self,
718        ddl_manager: DdlManagerRef,
719        procedure_manager: ProcedureManagerRef,
720    ) -> Result<ProcedureExecutorRef>;
721}
722
723pub struct DefaultProcedureExecutorCreator;
724
725#[async_trait]
726impl ProcedureExecutorCreator for DefaultProcedureExecutorCreator {
727    async fn create(
728        &self,
729        ddl_manager: DdlManagerRef,
730        procedure_manager: ProcedureManagerRef,
731    ) -> Result<ProcedureExecutorRef> {
732        Ok(Arc::new(LocalProcedureExecutor::new(
733            ddl_manager,
734            procedure_manager,
735        )))
736    }
737}
738
739#[async_trait]
740pub trait StandaloneLeaderServicesController: Send + Sync {
741    /// Starts services that manage standalone metadata or WAL state.
742    ///
743    /// The default implementation starts the procedure manager and WAL provider
744    /// during instance startup.
745    async fn start(
746        &self,
747        procedure_manager: ProcedureManagerRef,
748        wal_provider: WalProviderRef,
749        region_server: RegionServer,
750    ) -> Result<()>;
751
752    /// Stops services started by [`StandaloneLeaderServicesController::start`].
753    async fn stop(
754        &self,
755        procedure_manager: ProcedureManagerRef,
756        region_server: RegionServer,
757    ) -> Result<()>;
758}
759
760pub struct DefaultStandaloneLeaderServicesController;
761
762#[async_trait]
763impl StandaloneLeaderServicesController for DefaultStandaloneLeaderServicesController {
764    async fn start(
765        &self,
766        procedure_manager: ProcedureManagerRef,
767        wal_provider: WalProviderRef,
768        _region_server: RegionServer,
769    ) -> Result<()> {
770        procedure_manager
771            .start()
772            .await
773            .context(error::StartProcedureManagerSnafu)?;
774        wal_provider
775            .start()
776            .await
777            .context(error::StartWalProviderSnafu)
778    }
779
780    async fn stop(
781        &self,
782        procedure_manager: ProcedureManagerRef,
783        _region_server: RegionServer,
784    ) -> Result<()> {
785        procedure_manager
786            .stop()
787            .await
788            .context(error::StopProcedureManagerSnafu)
789    }
790}
791
792/// `InstanceCreator` is used for grouping various component creators for building the
793/// Standalone instance, suitable for customizing how the instance can be built.
794pub struct InstanceCreator {
795    /// Hook for replacing metadata KV construction while reusing the rest of the
796    /// standalone build flow.
797    metadata_kv_backend_creator: Box<dyn MetadataKvBackendCreator>,
798    node_manager_creator: Box<dyn NodeManagerCreator>,
799    table_id_allocator_creator: Box<dyn TableIdAllocatorCreator>,
800    procedure_executor_creator: Box<dyn ProcedureExecutorCreator>,
801    leader_services_controller: Box<dyn StandaloneLeaderServicesController>,
802    open_regions_writable_override: Option<bool>,
803}
804
805impl InstanceCreator {
806    pub fn new(
807        node_manager_creator: Box<dyn NodeManagerCreator>,
808        table_id_allocator_creator: Box<dyn TableIdAllocatorCreator>,
809        procedure_executor_creator: Box<dyn ProcedureExecutorCreator>,
810    ) -> Self {
811        Self {
812            metadata_kv_backend_creator: Box::new(DefaultMetadataKvBackendCreator),
813            node_manager_creator,
814            table_id_allocator_creator,
815            procedure_executor_creator,
816            leader_services_controller: Box::new(DefaultStandaloneLeaderServicesController),
817            open_regions_writable_override: None,
818        }
819    }
820
821    pub fn with_metadata_kv_backend_creator(
822        mut self,
823        metadata_kv_backend_creator: Box<dyn MetadataKvBackendCreator>,
824    ) -> Self {
825        self.metadata_kv_backend_creator = metadata_kv_backend_creator;
826        self
827    }
828
829    /// Wraps the metadata backend creator while retaining the default creator.
830    ///
831    /// This is useful for callers that need to add runtime behavior around
832    /// metadata access without reimplementing backend selection.
833    pub fn map_metadata_kv_backend_creator<F>(mut self, f: F) -> Self
834    where
835        F: FnOnce(Box<dyn MetadataKvBackendCreator>) -> Box<dyn MetadataKvBackendCreator>,
836    {
837        self.metadata_kv_backend_creator = f(self.metadata_kv_backend_creator);
838        self
839    }
840
841    /// Wraps node-manager creation while preserving the selected standalone node manager.
842    pub fn map_node_manager_creator<F>(mut self, f: F) -> Self
843    where
844        F: FnOnce(Box<dyn NodeManagerCreator>) -> Box<dyn NodeManagerCreator>,
845    {
846        self.node_manager_creator = f(self.node_manager_creator);
847        self
848    }
849
850    /// Wraps procedure-executor creation while preserving the current setup.
851    pub fn map_procedure_executor_creator<F>(mut self, f: F) -> Self
852    where
853        F: FnOnce(Box<dyn ProcedureExecutorCreator>) -> Box<dyn ProcedureExecutorCreator>,
854    {
855        self.procedure_executor_creator = f(self.procedure_executor_creator);
856        self
857    }
858
859    /// Replaces startup/shutdown ownership for procedure manager and WAL provider.
860    pub fn with_leader_services_controller(
861        mut self,
862        leader_services_controller: Box<dyn StandaloneLeaderServicesController>,
863    ) -> Self {
864        self.leader_services_controller = leader_services_controller;
865        self
866    }
867
868    /// Overrides whether regions opened during startup should become writable.
869    ///
870    /// `None` keeps the default startup behavior (regions open writable).
871    ///
872    /// Warning: setting this to `false` in standalone mode will leave reopened regions
873    /// permanently read-only. Standalone has no metasrv heartbeat or region-role
874    /// reconciliation, so there is no path to promote regions to Leader after startup.
875    pub fn with_open_regions_writable_override(mut self, writable: bool) -> Self {
876        self.open_regions_writable_override = Some(writable);
877        self
878    }
879}
880
881impl Default for InstanceCreator {
882    fn default() -> Self {
883        Self {
884            metadata_kv_backend_creator: Box::new(DefaultMetadataKvBackendCreator),
885            node_manager_creator: Box::new(DefaultNodeManagerCreator),
886            table_id_allocator_creator: Box::new(DefaultTableIdAllocatorCreator),
887            procedure_executor_creator: Box::new(DefaultProcedureExecutorCreator),
888            leader_services_controller: Box::new(DefaultStandaloneLeaderServicesController),
889            open_regions_writable_override: None,
890        }
891    }
892}
893
894/// `InstanceCreatorResult` is expected to be used paired with [InstanceCreator].
895/// It stores the created and other important components for further reusing.
896pub struct InstanceCreatorResult {
897    pub kv_backend: KvBackendRef,
898    pub node_manager: NodeManagerRef,
899    pub table_id_allocator: Arc<Sequence>,
900}
901
902#[cfg(test)]
903mod tests {
904    use std::default::Default;
905    use std::io::Write;
906    use std::time::Duration;
907
908    use auth::{Identity, Password, UserProviderRef};
909    use clap::{CommandFactory, Parser};
910    use common_base::readable_size::ReadableSize;
911    use common_config::ENV_VAR_SEP;
912    use common_options::plugin_options::StandaloneFlag;
913    use common_test_util::temp_dir::create_named_temp_file;
914    use common_wal::config::DatanodeWalConfig;
915    use frontend::frontend::FrontendOptions;
916    use object_store::config::{FileConfig, GcsConfig};
917    use servers::grpc::GrpcOptions;
918
919    use super::*;
920    use crate::options::GlobalOptions;
921
922    #[tokio::test]
923    async fn test_try_from_start_command_to_anymap() {
924        let fe_opts = FrontendOptions {
925            user_provider: Some("static_user_provider:cmd:test=test".to_string()),
926            ..Default::default()
927        };
928
929        let mut plugins = Plugins::new();
930        plugins.insert(StandaloneFlag);
931        plugins::setup_frontend_plugins(&mut plugins, &[], &fe_opts)
932            .await
933            .unwrap();
934
935        let provider = plugins.get::<UserProviderRef>().unwrap();
936        let result = provider
937            .authenticate(
938                Identity::UserId("test", None),
939                Password::PlainText("test".to_string().into()),
940            )
941            .await;
942        let _ = result.unwrap();
943    }
944
945    #[test]
946    fn test_toml() {
947        let opts = StandaloneOptions::default();
948        let toml_string = toml::to_string(&opts).unwrap();
949        let _parsed: StandaloneOptions = toml::from_str(&toml_string).unwrap();
950    }
951
952    #[test]
953    fn test_read_from_config_file() {
954        let mut file = create_named_temp_file();
955        let toml_str = r#"
956            enable_memory_catalog = true
957
958            [wal]
959            provider = "raft_engine"
960            dir = "./greptimedb_data/test/wal"
961            file_size = "1GB"
962            purge_threshold = "50GB"
963            purge_interval = "10m"
964            read_batch_size = 128
965            sync_write = false
966
967            [storage]
968            data_home = "./greptimedb_data/"
969            type = "File"
970
971            [[storage.providers]]
972            type = "Gcs"
973            bucket = "foo"
974            endpoint = "bar"
975
976            [[storage.providers]]
977            type = "S3"
978            access_key_id = "access_key_id"
979            secret_access_key = "secret_access_key"
980
981            [storage.compaction]
982            max_inflight_tasks = 3
983            max_files_in_level0 = 7
984            max_purge_tasks = 32
985
986            [storage.manifest]
987            checkpoint_margin = 9
988            gc_duration = '7s'
989
990            [http]
991            addr = "127.0.0.1:4000"
992            timeout = "33s"
993            body_limit = "128MB"
994
995            [opentsdb]
996            enable = true
997
998            [logging]
999            level = "debug"
1000            dir = "./greptimedb_data/test/logs"
1001        "#;
1002        write!(file, "{}", toml_str).unwrap();
1003        let cmd = StartCommand {
1004            config_file: Some(file.path().to_str().unwrap().to_string()),
1005            user_provider: Some("static_user_provider:cmd:test=test".to_string()),
1006            ..Default::default()
1007        };
1008
1009        let options = cmd
1010            .load_options(&GlobalOptions::default())
1011            .unwrap()
1012            .component;
1013        let fe_opts = options.frontend_options();
1014        let dn_opts = options.datanode_options();
1015        let logging_opts = options.logging;
1016        assert_eq!("127.0.0.1:4000".to_string(), fe_opts.http.addr);
1017        assert_eq!(Duration::from_secs(33), fe_opts.http.timeout);
1018        assert_eq!(ReadableSize::mb(128), fe_opts.http.body_limit);
1019        assert_eq!("127.0.0.1:4001".to_string(), fe_opts.grpc.bind_addr);
1020        assert!(fe_opts.mysql.enable);
1021        assert_eq!("127.0.0.1:4002", fe_opts.mysql.addr);
1022        assert_eq!(2, fe_opts.mysql.runtime_size);
1023        assert_eq!(None, fe_opts.mysql.reject_no_database);
1024        assert!(fe_opts.influxdb.enable);
1025        assert!(fe_opts.opentsdb.enable);
1026
1027        let DatanodeWalConfig::RaftEngine(raft_engine_config) = dn_opts.wal else {
1028            unreachable!()
1029        };
1030        assert_eq!(
1031            "./greptimedb_data/test/wal",
1032            raft_engine_config.dir.unwrap()
1033        );
1034
1035        assert!(matches!(
1036            &dn_opts.storage.store,
1037            object_store::config::ObjectStoreConfig::File(FileConfig { .. })
1038        ));
1039        assert_eq!(dn_opts.storage.providers.len(), 2);
1040        assert!(matches!(
1041            dn_opts.storage.providers[0],
1042            object_store::config::ObjectStoreConfig::Gcs(GcsConfig { .. })
1043        ));
1044        match &dn_opts.storage.providers[1] {
1045            object_store::config::ObjectStoreConfig::S3(s3_config) => {
1046                assert_eq!(
1047                    "SecretBox<alloc::string::String>([REDACTED])".to_string(),
1048                    format!("{:?}", s3_config.connection.access_key_id)
1049                );
1050            }
1051            _ => {
1052                unreachable!()
1053            }
1054        }
1055
1056        assert_eq!("debug", logging_opts.level.as_ref().unwrap());
1057        assert_eq!("./greptimedb_data/test/logs".to_string(), logging_opts.dir);
1058    }
1059
1060    #[test]
1061    fn test_load_log_options_from_cli() {
1062        let cmd = StartCommand {
1063            user_provider: Some("static_user_provider:cmd:test=test".to_string()),
1064            mysql_addr: Some("127.0.0.1:4002".to_string()),
1065            postgres_addr: Some("127.0.0.1:4003".to_string()),
1066            ..Default::default()
1067        };
1068
1069        let opts = cmd
1070            .load_options(&GlobalOptions {
1071                log_dir: Some("./greptimedb_data/test/logs".to_string()),
1072                log_level: Some("debug".to_string()),
1073
1074                #[cfg(feature = "tokio-console")]
1075                tokio_console_addr: None,
1076            })
1077            .unwrap()
1078            .component;
1079
1080        assert_eq!("./greptimedb_data/test/logs", opts.logging.dir);
1081        assert_eq!("debug", opts.logging.level.unwrap());
1082    }
1083
1084    #[test]
1085    fn test_config_precedence_order() {
1086        let mut file = create_named_temp_file();
1087        let toml_str = r#"
1088            [http]
1089            addr = "127.0.0.1:4000"
1090
1091            [logging]
1092            level = "debug"
1093        "#;
1094        write!(file, "{}", toml_str).unwrap();
1095
1096        let env_prefix = "STANDALONE_UT";
1097        temp_env::with_vars(
1098            [
1099                (
1100                    // logging.dir = /other/log/dir
1101                    [
1102                        env_prefix.to_string(),
1103                        "logging".to_uppercase(),
1104                        "dir".to_uppercase(),
1105                    ]
1106                    .join(ENV_VAR_SEP),
1107                    Some("/other/log/dir"),
1108                ),
1109                (
1110                    // logging.level = info
1111                    [
1112                        env_prefix.to_string(),
1113                        "logging".to_uppercase(),
1114                        "level".to_uppercase(),
1115                    ]
1116                    .join(ENV_VAR_SEP),
1117                    Some("info"),
1118                ),
1119                (
1120                    // http.addr = 127.0.0.1:24000
1121                    [
1122                        env_prefix.to_string(),
1123                        "http".to_uppercase(),
1124                        "addr".to_uppercase(),
1125                    ]
1126                    .join(ENV_VAR_SEP),
1127                    Some("127.0.0.1:24000"),
1128                ),
1129            ],
1130            || {
1131                let command = StartCommand {
1132                    config_file: Some(file.path().to_str().unwrap().to_string()),
1133                    http_addr: Some("127.0.0.1:14000".to_string()),
1134                    env_prefix: env_prefix.to_string(),
1135                    ..Default::default()
1136                };
1137
1138                let opts = command.load_options(&Default::default()).unwrap().component;
1139
1140                // Should be read from env, env > default values.
1141                assert_eq!(opts.logging.dir, "/other/log/dir");
1142
1143                // Should be read from config file, config file > env > default values.
1144                assert_eq!(opts.logging.level.as_ref().unwrap(), "debug");
1145
1146                // Should be read from cli, cli > config file > env > default values.
1147                let fe_opts = opts.frontend_options();
1148                assert_eq!(fe_opts.http.addr, "127.0.0.1:14000");
1149                assert_eq!(ReadableSize::mb(64), fe_opts.http.body_limit);
1150
1151                // Should be default value.
1152                assert_eq!(fe_opts.grpc.bind_addr, GrpcOptions::default().bind_addr);
1153            },
1154        );
1155    }
1156
1157    #[test]
1158    fn test_parse_grpc_bind_addr_aliases() {
1159        let command =
1160            StartCommand::try_parse_from(["standalone", "--grpc-bind-addr", "127.0.0.1:14001"])
1161                .unwrap();
1162        assert_eq!(command.grpc_bind_addr.as_deref(), Some("127.0.0.1:14001"));
1163
1164        let command =
1165            StartCommand::try_parse_from(["standalone", "--rpc-bind-addr", "127.0.0.1:24001"])
1166                .unwrap();
1167        assert_eq!(command.grpc_bind_addr.as_deref(), Some("127.0.0.1:24001"));
1168
1169        let command =
1170            StartCommand::try_parse_from(["standalone", "--rpc-addr", "127.0.0.1:34001"]).unwrap();
1171        assert_eq!(command.grpc_bind_addr.as_deref(), Some("127.0.0.1:34001"));
1172    }
1173
1174    #[test]
1175    fn test_help_uses_grpc_option_names() {
1176        let mut cmd = StartCommand::command();
1177        let mut help = Vec::new();
1178        cmd.write_long_help(&mut help).unwrap();
1179        let help = String::from_utf8(help).unwrap();
1180
1181        assert!(help.contains("--grpc-bind-addr"));
1182        assert!(!help.contains("--rpc-bind-addr"));
1183        assert!(!help.contains("--rpc-addr"));
1184    }
1185
1186    #[test]
1187    fn test_load_default_standalone_options() {
1188        let options =
1189            StandaloneOptions::load_layered_options(None, "GREPTIMEDB_STANDALONE").unwrap();
1190        let default_options = StandaloneOptions::default();
1191        assert_eq!(options.enable_telemetry, default_options.enable_telemetry);
1192        assert_eq!(options.http, default_options.http);
1193        assert_eq!(options.grpc, default_options.grpc);
1194        assert_eq!(options.mysql, default_options.mysql);
1195        assert_eq!(options.postgres, default_options.postgres);
1196        assert_eq!(options.opentsdb, default_options.opentsdb);
1197        assert_eq!(options.influxdb, default_options.influxdb);
1198        assert_eq!(options.prom_store, default_options.prom_store);
1199        assert_eq!(options.wal, default_options.wal);
1200        assert_eq!(options.metadata_store, default_options.metadata_store);
1201        assert_eq!(options.procedure, default_options.procedure);
1202        assert_eq!(options.logging, default_options.logging);
1203        assert_eq!(options.region_engine, default_options.region_engine);
1204    }
1205
1206    #[test]
1207    fn test_cache_config() {
1208        let toml_str = r#"
1209            [storage]
1210            data_home = "test_data_home"
1211            type = "S3"
1212            [storage.cache_config]
1213            enable_read_cache = true
1214        "#;
1215        let mut opts: StandaloneOptions = toml::from_str(toml_str).unwrap();
1216        opts.sanitize();
1217        assert!(opts.storage.store.cache_config().unwrap().enable_read_cache);
1218        assert_eq!(
1219            opts.storage.store.cache_config().unwrap().cache_path,
1220            "test_data_home"
1221        );
1222    }
1223}