Separate Python and Rust into python/ and rust/ with per-stack Dockerfiles

rust/myfsio-engine/crates/myfsio-server/src/services/gc.rs

@@ -0,0 +1,263 @@
+use serde_json::{json, Value};
+use std::path::PathBuf;
+use std::sync::Arc;
+use std::time::Instant;
+use tokio::sync::RwLock;
+
+pub struct GcConfig {
+    pub interval_hours: f64,
+    pub temp_file_max_age_hours: f64,
+    pub multipart_max_age_days: u64,
+    pub lock_file_max_age_hours: f64,
+    pub dry_run: bool,
+}
+
+impl Default for GcConfig {
+    fn default() -> Self {
+        Self {
+            interval_hours: 6.0,
+            temp_file_max_age_hours: 24.0,
+            multipart_max_age_days: 7,
+            lock_file_max_age_hours: 1.0,
+            dry_run: false,
+        }
+    }
+}
+
+pub struct GcService {
+    storage_root: PathBuf,
+    config: GcConfig,
+    running: Arc<RwLock<bool>>,
+    history: Arc<RwLock<Vec<Value>>>,
+    history_path: PathBuf,
+}
+
+impl GcService {
+    pub fn new(storage_root: PathBuf, config: GcConfig) -> Self {
+        let history_path = storage_root
+            .join(".myfsio.sys")
+            .join("config")
+            .join("gc_history.json");
+
+        let history = if history_path.exists() {
+            std::fs::read_to_string(&history_path)
+                .ok()
+                .and_then(|s| serde_json::from_str::<Value>(&s).ok())
+                .and_then(|v| v.get("executions").and_then(|e| e.as_array().cloned()))
+                .unwrap_or_default()
+        } else {
+            Vec::new()
+        };
+
+        Self {
+            storage_root,
+            config,
+            running: Arc::new(RwLock::new(false)),
+            history: Arc::new(RwLock::new(history)),
+            history_path,
+        }
+    }
+
+    pub async fn status(&self) -> Value {
+        let running = *self.running.read().await;
+        json!({
+            "enabled": true,
+            "running": running,
+            "interval_hours": self.config.interval_hours,
+            "temp_file_max_age_hours": self.config.temp_file_max_age_hours,
+            "multipart_max_age_days": self.config.multipart_max_age_days,
+            "lock_file_max_age_hours": self.config.lock_file_max_age_hours,
+            "dry_run": self.config.dry_run,
+        })
+    }
+
+    pub async fn history(&self) -> Value {
+        let history = self.history.read().await;
+        json!({ "executions": *history })
+    }
+
+    pub async fn run_now(&self, dry_run: bool) -> Result<Value, String> {
+        {
+            let mut running = self.running.write().await;
+            if *running {
+                return Err("GC already running".to_string());
+            }
+            *running = true;
+        }
+
+        let start = Instant::now();
+        let result = self.execute_gc(dry_run || self.config.dry_run).await;
+        let elapsed = start.elapsed().as_secs_f64();
+
+        *self.running.write().await = false;
+
+        let mut result_json = result.clone();
+        if let Some(obj) = result_json.as_object_mut() {
+            obj.insert("execution_time_seconds".to_string(), json!(elapsed));
+        }
+
+        let record = json!({
+            "timestamp": chrono::Utc::now().timestamp_millis() as f64 / 1000.0,
+            "dry_run": dry_run || self.config.dry_run,
+            "result": result_json,
+        });
+
+        {
+            let mut history = self.history.write().await;
+            history.push(record);
+            if history.len() > 50 {
+                let excess = history.len() - 50;
+                history.drain(..excess);
+            }
+        }
+        self.save_history().await;
+
+        Ok(result)
+    }
+
+    async fn execute_gc(&self, dry_run: bool) -> Value {
+        let mut temp_files_deleted = 0u64;
+        let mut temp_bytes_freed = 0u64;
+        let mut multipart_uploads_deleted = 0u64;
+        let mut lock_files_deleted = 0u64;
+        let mut empty_dirs_removed = 0u64;
+        let mut errors: Vec<String> = Vec::new();
+
+        let now = std::time::SystemTime::now();
+        let temp_max_age = std::time::Duration::from_secs_f64(self.config.temp_file_max_age_hours * 3600.0);
+        let multipart_max_age = std::time::Duration::from_secs(self.config.multipart_max_age_days * 86400);
+        let lock_max_age = std::time::Duration::from_secs_f64(self.config.lock_file_max_age_hours * 3600.0);
+
+        let tmp_dir = self.storage_root.join(".myfsio.sys").join("tmp");
+        if tmp_dir.exists() {
+            match std::fs::read_dir(&tmp_dir) {
+                Ok(entries) => {
+                    for entry in entries.flatten() {
+                        if let Ok(metadata) = entry.metadata() {
+                            if let Ok(modified) = metadata.modified() {
+                                if let Ok(age) = now.duration_since(modified) {
+                                    if age > temp_max_age {
+                                        let size = metadata.len();
+                                        if !dry_run {
+                                            if let Err(e) = std::fs::remove_file(entry.path()) {
+                                                errors.push(format!("Failed to remove temp file: {}", e));
+                                                continue;
+                                            }
+                                        }
+                                        temp_files_deleted += 1;
+                                        temp_bytes_freed += size;
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+                Err(e) => errors.push(format!("Failed to read tmp dir: {}", e)),
+            }
+        }
+
+        let multipart_dir = self.storage_root.join(".myfsio.sys").join("multipart");
+        if multipart_dir.exists() {
+            if let Ok(bucket_dirs) = std::fs::read_dir(&multipart_dir) {
+                for bucket_entry in bucket_dirs.flatten() {
+                    if let Ok(uploads) = std::fs::read_dir(bucket_entry.path()) {
+                        for upload in uploads.flatten() {
+                            if let Ok(metadata) = upload.metadata() {
+                                if let Ok(modified) = metadata.modified() {
+                                    if let Ok(age) = now.duration_since(modified) {
+                                        if age > multipart_max_age {
+                                            if !dry_run {
+                                                let _ = std::fs::remove_dir_all(upload.path());
+                                            }
+                                            multipart_uploads_deleted += 1;
+                                        }
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
+        let buckets_dir = self.storage_root.join(".myfsio.sys").join("buckets");
+        if buckets_dir.exists() {
+            if let Ok(bucket_dirs) = std::fs::read_dir(&buckets_dir) {
+                for bucket_entry in bucket_dirs.flatten() {
+                    let locks_dir = bucket_entry.path().join("locks");
+                    if locks_dir.exists() {
+                        if let Ok(locks) = std::fs::read_dir(&locks_dir) {
+                            for lock in locks.flatten() {
+                                if let Ok(metadata) = lock.metadata() {
+                                    if let Ok(modified) = metadata.modified() {
+                                        if let Ok(age) = now.duration_since(modified) {
+                                            if age > lock_max_age {
+                                                if !dry_run {
+                                                    let _ = std::fs::remove_file(lock.path());
+                                                }
+                                                lock_files_deleted += 1;
+                                            }
+                                        }
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
+        if !dry_run {
+            for dir in [&tmp_dir, &multipart_dir] {
+                if dir.exists() {
+                    if let Ok(entries) = std::fs::read_dir(dir) {
+                        for entry in entries.flatten() {
+                            if entry.path().is_dir() {
+                                if let Ok(mut contents) = std::fs::read_dir(entry.path()) {
+                                    if contents.next().is_none() {
+                                        let _ = std::fs::remove_dir(entry.path());
+                                        empty_dirs_removed += 1;
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
+        json!({
+            "temp_files_deleted": temp_files_deleted,
+            "temp_bytes_freed": temp_bytes_freed,
+            "multipart_uploads_deleted": multipart_uploads_deleted,
+            "lock_files_deleted": lock_files_deleted,
+            "empty_dirs_removed": empty_dirs_removed,
+            "errors": errors,
+        })
+    }
+
+    async fn save_history(&self) {
+        let history = self.history.read().await;
+        let data = json!({ "executions": *history });
+        if let Some(parent) = self.history_path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        let _ = std::fs::write(&self.history_path, serde_json::to_string_pretty(&data).unwrap_or_default());
+    }
+
+    pub fn start_background(self: Arc<Self>) -> tokio::task::JoinHandle<()> {
+        let interval = std::time::Duration::from_secs_f64(self.config.interval_hours * 3600.0);
+        tokio::spawn(async move {
+            let mut timer = tokio::time::interval(interval);
+            timer.tick().await;
+            loop {
+                timer.tick().await;
+                tracing::info!("GC cycle starting");
+                match self.run_now(false).await {
+                    Ok(result) => tracing::info!("GC cycle complete: {:?}", result),
+                    Err(e) => tracing::warn!("GC cycle failed: {}", e),
+                }
+            }
+        })
+    }
+}

rust/myfsio-engine/crates/myfsio-server/src/services/integrity.rs

@@ -0,0 +1,204 @@
+use myfsio_storage::fs_backend::FsStorageBackend;
+use myfsio_storage::traits::StorageEngine;
+use serde_json::{json, Value};
+use std::path::PathBuf;
+use std::sync::Arc;
+use std::time::Instant;
+use tokio::sync::RwLock;
+
+pub struct IntegrityConfig {
+    pub interval_hours: f64,
+    pub batch_size: usize,
+    pub auto_heal: bool,
+    pub dry_run: bool,
+}
+
+impl Default for IntegrityConfig {
+    fn default() -> Self {
+        Self {
+            interval_hours: 24.0,
+            batch_size: 1000,
+            auto_heal: false,
+            dry_run: false,
+        }
+    }
+}
+
+pub struct IntegrityService {
+    storage: Arc<FsStorageBackend>,
+    config: IntegrityConfig,
+    running: Arc<RwLock<bool>>,
+    history: Arc<RwLock<Vec<Value>>>,
+    history_path: PathBuf,
+}
+
+impl IntegrityService {
+    pub fn new(
+        storage: Arc<FsStorageBackend>,
+        storage_root: &std::path::Path,
+        config: IntegrityConfig,
+    ) -> Self {
+        let history_path = storage_root
+            .join(".myfsio.sys")
+            .join("config")
+            .join("integrity_history.json");
+
+        let history = if history_path.exists() {
+            std::fs::read_to_string(&history_path)
+                .ok()
+                .and_then(|s| serde_json::from_str::<Value>(&s).ok())
+                .and_then(|v| v.get("executions").and_then(|e| e.as_array().cloned()))
+                .unwrap_or_default()
+        } else {
+            Vec::new()
+        };
+
+        Self {
+            storage,
+            config,
+            running: Arc::new(RwLock::new(false)),
+            history: Arc::new(RwLock::new(history)),
+            history_path,
+        }
+    }
+
+    pub async fn status(&self) -> Value {
+        let running = *self.running.read().await;
+        json!({
+            "enabled": true,
+            "running": running,
+            "interval_hours": self.config.interval_hours,
+            "batch_size": self.config.batch_size,
+            "auto_heal": self.config.auto_heal,
+            "dry_run": self.config.dry_run,
+        })
+    }
+
+    pub async fn history(&self) -> Value {
+        let history = self.history.read().await;
+        json!({ "executions": *history })
+    }
+
+    pub async fn run_now(&self, dry_run: bool, auto_heal: bool) -> Result<Value, String> {
+        {
+            let mut running = self.running.write().await;
+            if *running {
+                return Err("Integrity check already running".to_string());
+            }
+            *running = true;
+        }
+
+        let start = Instant::now();
+        let result = self.check_integrity(dry_run, auto_heal).await;
+        let elapsed = start.elapsed().as_secs_f64();
+
+        *self.running.write().await = false;
+
+        let mut result_json = result.clone();
+        if let Some(obj) = result_json.as_object_mut() {
+            obj.insert("execution_time_seconds".to_string(), json!(elapsed));
+        }
+
+        let record = json!({
+            "timestamp": chrono::Utc::now().timestamp_millis() as f64 / 1000.0,
+            "dry_run": dry_run,
+            "auto_heal": auto_heal,
+            "result": result_json,
+        });
+
+        {
+            let mut history = self.history.write().await;
+            history.push(record);
+            if history.len() > 50 {
+                let excess = history.len() - 50;
+                history.drain(..excess);
+            }
+        }
+        self.save_history().await;
+
+        Ok(result)
+    }
+
+    async fn check_integrity(&self, _dry_run: bool, _auto_heal: bool) -> Value {
+        let buckets = match self.storage.list_buckets().await {
+            Ok(b) => b,
+            Err(e) => return json!({"error": e.to_string()}),
+        };
+
+        let mut objects_scanned = 0u64;
+        let mut corrupted = 0u64;
+        let mut phantom_metadata = 0u64;
+        let mut errors: Vec<String> = Vec::new();
+
+        for bucket in &buckets {
+            let params = myfsio_common::types::ListParams {
+                max_keys: self.config.batch_size,
+                ..Default::default()
+            };
+            let objects = match self.storage.list_objects(&bucket.name, &params).await {
+                Ok(r) => r.objects,
+                Err(e) => {
+                    errors.push(format!("{}: {}", bucket.name, e));
+                    continue;
+                }
+            };
+
+            for obj in &objects {
+                objects_scanned += 1;
+                match self.storage.get_object_path(&bucket.name, &obj.key).await {
+                    Ok(path) => {
+                        if !path.exists() {
+                            phantom_metadata += 1;
+                        } else if let Some(ref expected_etag) = obj.etag {
+                            match myfsio_crypto::hashing::md5_file(&path) {
+                                Ok(actual_etag) => {
+                                    if &actual_etag != expected_etag {
+                                        corrupted += 1;
+                                    }
+                                }
+                                Err(e) => errors.push(format!("{}:{}: {}", bucket.name, obj.key, e)),
+                            }
+                        }
+                    }
+                    Err(e) => errors.push(format!("{}:{}: {}", bucket.name, obj.key, e)),
+                }
+            }
+        }
+
+        json!({
+            "objects_scanned": objects_scanned,
+            "buckets_scanned": buckets.len(),
+            "corrupted_objects": corrupted,
+            "phantom_metadata": phantom_metadata,
+            "errors": errors,
+        })
+    }
+
+    async fn save_history(&self) {
+        let history = self.history.read().await;
+        let data = json!({ "executions": *history });
+        if let Some(parent) = self.history_path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        let _ = std::fs::write(
+            &self.history_path,
+            serde_json::to_string_pretty(&data).unwrap_or_default(),
+        );
+    }
+
+    pub fn start_background(self: Arc<Self>) -> tokio::task::JoinHandle<()> {
+        let interval = std::time::Duration::from_secs_f64(self.config.interval_hours * 3600.0);
+        tokio::spawn(async move {
+            let mut timer = tokio::time::interval(interval);
+            timer.tick().await;
+            loop {
+                timer.tick().await;
+                tracing::info!("Integrity check starting");
+                match self.run_now(false, false).await {
+                    Ok(result) => tracing::info!("Integrity check complete: {:?}", result),
+                    Err(e) => tracing::warn!("Integrity check failed: {}", e),
+                }
+            }
+        })
+    }
+}

rust/myfsio-engine/crates/myfsio-server/src/services/lifecycle.rs

@@ -0,0 +1,153 @@
+use myfsio_storage::fs_backend::FsStorageBackend;
+use myfsio_storage::traits::StorageEngine;
+use serde_json::{json, Value};
+use std::sync::Arc;
+use tokio::sync::RwLock;
+
+pub struct LifecycleConfig {
+    pub interval_seconds: u64,
+}
+
+impl Default for LifecycleConfig {
+    fn default() -> Self {
+        Self {
+            interval_seconds: 3600,
+        }
+    }
+}
+
+pub struct LifecycleService {
+    storage: Arc<FsStorageBackend>,
+    config: LifecycleConfig,
+    running: Arc<RwLock<bool>>,
+}
+
+impl LifecycleService {
+    pub fn new(storage: Arc<FsStorageBackend>, config: LifecycleConfig) -> Self {
+        Self {
+            storage,
+            config,
+            running: Arc::new(RwLock::new(false)),
+        }
+    }
+
+    pub async fn run_cycle(&self) -> Result<Value, String> {
+        {
+            let mut running = self.running.write().await;
+            if *running {
+                return Err("Lifecycle already running".to_string());
+            }
+            *running = true;
+        }
+
+        let result = self.evaluate_rules().await;
+        *self.running.write().await = false;
+        Ok(result)
+    }
+
+    async fn evaluate_rules(&self) -> Value {
+        let buckets = match self.storage.list_buckets().await {
+            Ok(b) => b,
+            Err(e) => return json!({"error": e.to_string()}),
+        };
+
+        let mut total_expired = 0u64;
+        let mut total_multipart_aborted = 0u64;
+        let mut errors: Vec<String> = Vec::new();
+
+        for bucket in &buckets {
+            let config = match self.storage.get_bucket_config(&bucket.name).await {
+                Ok(c) => c,
+                Err(_) => continue,
+            };
+
+            let lifecycle = match &config.lifecycle {
+                Some(lc) => lc,
+                None => continue,
+            };
+
+            let rules = match lifecycle.as_str().and_then(|s| serde_json::from_str::<Value>(s).ok()) {
+                Some(v) => v,
+                None => continue,
+            };
+
+            let rules_arr = match rules.get("Rules").and_then(|r| r.as_array()) {
+                Some(a) => a.clone(),
+                None => continue,
+            };
+
+            for rule in &rules_arr {
+                if rule.get("Status").and_then(|s| s.as_str()) != Some("Enabled") {
+                    continue;
+                }
+
+                let prefix = rule
+                    .get("Filter")
+                    .and_then(|f| f.get("Prefix"))
+                    .and_then(|p| p.as_str())
+                    .or_else(|| rule.get("Prefix").and_then(|p| p.as_str()))
+                    .unwrap_or("");
+
+                if let Some(exp) = rule.get("Expiration") {
+                    if let Some(days) = exp.get("Days").and_then(|d| d.as_u64()) {
+                        let cutoff = chrono::Utc::now() - chrono::Duration::days(days as i64);
+                        let params = myfsio_common::types::ListParams {
+                            max_keys: 1000,
+                            prefix: if prefix.is_empty() { None } else { Some(prefix.to_string()) },
+                            ..Default::default()
+                        };
+                        if let Ok(result) = self.storage.list_objects(&bucket.name, &params).await {
+                            for obj in &result.objects {
+                                if obj.last_modified < cutoff {
+                                    match self.storage.delete_object(&bucket.name, &obj.key).await {
+                                        Ok(()) => total_expired += 1,
+                                        Err(e) => errors.push(format!("{}:{}: {}", bucket.name, obj.key, e)),
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+
+                if let Some(abort) = rule.get("AbortIncompleteMultipartUpload") {
+                    if let Some(days) = abort.get("DaysAfterInitiation").and_then(|d| d.as_u64()) {
+                        let cutoff = chrono::Utc::now() - chrono::Duration::days(days as i64);
+                        if let Ok(uploads) = self.storage.list_multipart_uploads(&bucket.name).await {
+                            for upload in &uploads {
+                                if upload.initiated < cutoff {
+                                    match self.storage.abort_multipart(&bucket.name, &upload.upload_id).await {
+                                        Ok(()) => total_multipart_aborted += 1,
+                                        Err(e) => errors.push(format!("abort {}: {}", upload.upload_id, e)),
+                                    }
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+        }
+
+        json!({
+            "objects_expired": total_expired,
+            "multipart_aborted": total_multipart_aborted,
+            "buckets_evaluated": buckets.len(),
+            "errors": errors,
+        })
+    }
+
+    pub fn start_background(self: Arc<Self>) -> tokio::task::JoinHandle<()> {
+        let interval = std::time::Duration::from_secs(self.config.interval_seconds);
+        tokio::spawn(async move {
+            let mut timer = tokio::time::interval(interval);
+            timer.tick().await;
+            loop {
+                timer.tick().await;
+                tracing::info!("Lifecycle evaluation starting");
+                match self.run_cycle().await {
+                    Ok(result) => tracing::info!("Lifecycle cycle complete: {:?}", result),
+                    Err(e) => tracing::warn!("Lifecycle cycle failed: {}", e),
+                }
+            }
+        })
+    }
+}

rust/myfsio-engine/crates/myfsio-server/src/services/metrics.rs

@@ -0,0 +1,365 @@
+use chrono::{DateTime, Utc};
+use parking_lot::Mutex;
+use rand::Rng;
+use serde::{Deserialize, Serialize};
+use serde_json::{json, Value};
+use std::collections::HashMap;
+use std::path::{Path, PathBuf};
+use std::sync::Arc;
+use std::time::{SystemTime, UNIX_EPOCH};
+
+const MAX_LATENCY_SAMPLES: usize = 5000;
+
+pub struct MetricsConfig {
+    pub interval_minutes: u64,
+    pub retention_hours: u64,
+}
+
+impl Default for MetricsConfig {
+    fn default() -> Self {
+        Self {
+            interval_minutes: 5,
+            retention_hours: 24,
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+struct OperationStats {
+    count: u64,
+    success_count: u64,
+    error_count: u64,
+    latency_sum_ms: f64,
+    latency_min_ms: f64,
+    latency_max_ms: f64,
+    bytes_in: u64,
+    bytes_out: u64,
+    latency_samples: Vec<f64>,
+}
+
+impl Default for OperationStats {
+    fn default() -> Self {
+        Self {
+            count: 0,
+            success_count: 0,
+            error_count: 0,
+            latency_sum_ms: 0.0,
+            latency_min_ms: f64::INFINITY,
+            latency_max_ms: 0.0,
+            bytes_in: 0,
+            bytes_out: 0,
+            latency_samples: Vec::new(),
+        }
+    }
+}
+
+impl OperationStats {
+    fn record(&mut self, latency_ms: f64, success: bool, bytes_in: u64, bytes_out: u64) {
+        self.count += 1;
+        if success {
+            self.success_count += 1;
+        } else {
+            self.error_count += 1;
+        }
+        self.latency_sum_ms += latency_ms;
+        if latency_ms < self.latency_min_ms {
+            self.latency_min_ms = latency_ms;
+        }
+        if latency_ms > self.latency_max_ms {
+            self.latency_max_ms = latency_ms;
+        }
+        self.bytes_in += bytes_in;
+        self.bytes_out += bytes_out;
+
+        if self.latency_samples.len() < MAX_LATENCY_SAMPLES {
+            self.latency_samples.push(latency_ms);
+        } else {
+            let mut rng = rand::thread_rng();
+            let j = rng.gen_range(0..self.count as usize);
+            if j < MAX_LATENCY_SAMPLES {
+                self.latency_samples[j] = latency_ms;
+            }
+        }
+    }
+
+    fn compute_percentile(sorted: &[f64], p: f64) -> f64 {
+        if sorted.is_empty() {
+            return 0.0;
+        }
+        let k = (sorted.len() - 1) as f64 * (p / 100.0);
+        let f = k.floor() as usize;
+        let c = (f + 1).min(sorted.len() - 1);
+        let d = k - f as f64;
+        sorted[f] + d * (sorted[c] - sorted[f])
+    }
+
+    fn to_json(&self) -> Value {
+        let avg = if self.count > 0 {
+            self.latency_sum_ms / self.count as f64
+        } else {
+            0.0
+        };
+        let min = if self.latency_min_ms.is_infinite() {
+            0.0
+        } else {
+            self.latency_min_ms
+        };
+        let mut sorted = self.latency_samples.clone();
+        sorted.sort_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal));
+        json!({
+            "count": self.count,
+            "success_count": self.success_count,
+            "error_count": self.error_count,
+            "latency_avg_ms": round2(avg),
+            "latency_min_ms": round2(min),
+            "latency_max_ms": round2(self.latency_max_ms),
+            "latency_p50_ms": round2(Self::compute_percentile(&sorted, 50.0)),
+            "latency_p95_ms": round2(Self::compute_percentile(&sorted, 95.0)),
+            "latency_p99_ms": round2(Self::compute_percentile(&sorted, 99.0)),
+            "bytes_in": self.bytes_in,
+            "bytes_out": self.bytes_out,
+        })
+    }
+}
+
+fn round2(v: f64) -> f64 {
+    (v * 100.0).round() / 100.0
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct MetricsSnapshot {
+    pub timestamp: DateTime<Utc>,
+    pub window_seconds: u64,
+    pub by_method: HashMap<String, Value>,
+    pub by_endpoint: HashMap<String, Value>,
+    pub by_status_class: HashMap<String, u64>,
+    pub error_codes: HashMap<String, u64>,
+    pub totals: Value,
+}
+
+struct Inner {
+    by_method: HashMap<String, OperationStats>,
+    by_endpoint: HashMap<String, OperationStats>,
+    by_status_class: HashMap<String, u64>,
+    error_codes: HashMap<String, u64>,
+    totals: OperationStats,
+    window_start: f64,
+    snapshots: Vec<MetricsSnapshot>,
+}
+
+pub struct MetricsService {
+    config: MetricsConfig,
+    inner: Arc<Mutex<Inner>>,
+    snapshots_path: PathBuf,
+}
+
+impl MetricsService {
+    pub fn new(storage_root: &Path, config: MetricsConfig) -> Self {
+        let snapshots_path = storage_root
+            .join(".myfsio.sys")
+            .join("config")
+            .join("operation_metrics.json");
+
+        let mut snapshots: Vec<MetricsSnapshot> = if snapshots_path.exists() {
+            std::fs::read_to_string(&snapshots_path)
+                .ok()
+                .and_then(|s| serde_json::from_str::<Value>(&s).ok())
+                .and_then(|v| {
+                    v.get("snapshots")
+                        .and_then(|s| serde_json::from_value::<Vec<MetricsSnapshot>>(s.clone()).ok())
+                })
+                .unwrap_or_default()
+        } else {
+            Vec::new()
+        };
+        let cutoff = now_secs() - (config.retention_hours * 3600) as f64;
+        snapshots.retain(|s| s.timestamp.timestamp() as f64 > cutoff);
+
+        Self {
+            config,
+            inner: Arc::new(Mutex::new(Inner {
+                by_method: HashMap::new(),
+                by_endpoint: HashMap::new(),
+                by_status_class: HashMap::new(),
+                error_codes: HashMap::new(),
+                totals: OperationStats::default(),
+                window_start: now_secs(),
+                snapshots,
+            })),
+            snapshots_path,
+        }
+    }
+
+    pub fn record_request(
+        &self,
+        method: &str,
+        endpoint_type: &str,
+        status_code: u16,
+        latency_ms: f64,
+        bytes_in: u64,
+        bytes_out: u64,
+        error_code: Option<&str>,
+    ) {
+        let success = (200..400).contains(&status_code);
+        let status_class = format!("{}xx", status_code / 100);
+
+        let mut inner = self.inner.lock();
+        inner
+            .by_method
+            .entry(method.to_string())
+            .or_default()
+            .record(latency_ms, success, bytes_in, bytes_out);
+        inner
+            .by_endpoint
+            .entry(endpoint_type.to_string())
+            .or_default()
+            .record(latency_ms, success, bytes_in, bytes_out);
+        *inner.by_status_class.entry(status_class).or_insert(0) += 1;
+        if let Some(code) = error_code {
+            *inner.error_codes.entry(code.to_string()).or_insert(0) += 1;
+        }
+        inner.totals.record(latency_ms, success, bytes_in, bytes_out);
+    }
+
+    pub fn get_current_stats(&self) -> Value {
+        let inner = self.inner.lock();
+        let window_seconds = (now_secs() - inner.window_start).max(0.0) as u64;
+        let by_method: HashMap<String, Value> = inner
+            .by_method
+            .iter()
+            .map(|(k, v)| (k.clone(), v.to_json()))
+            .collect();
+        let by_endpoint: HashMap<String, Value> = inner
+            .by_endpoint
+            .iter()
+            .map(|(k, v)| (k.clone(), v.to_json()))
+            .collect();
+        json!({
+            "timestamp": Utc::now().to_rfc3339(),
+            "window_seconds": window_seconds,
+            "by_method": by_method,
+            "by_endpoint": by_endpoint,
+            "by_status_class": inner.by_status_class,
+            "error_codes": inner.error_codes,
+            "totals": inner.totals.to_json(),
+        })
+    }
+
+    pub fn get_history(&self, hours: Option<u64>) -> Vec<MetricsSnapshot> {
+        let inner = self.inner.lock();
+        let mut snapshots = inner.snapshots.clone();
+        if let Some(h) = hours {
+            let cutoff = now_secs() - (h * 3600) as f64;
+            snapshots.retain(|s| s.timestamp.timestamp() as f64 > cutoff);
+        }
+        snapshots
+    }
+
+    pub fn snapshot(&self) -> Value {
+        let current = self.get_current_stats();
+        let history = self.get_history(None);
+        json!({
+            "enabled": true,
+            "current": current,
+            "snapshots": history,
+        })
+    }
+
+    fn take_snapshot(&self) {
+        let snapshot = {
+            let mut inner = self.inner.lock();
+            let window_seconds = (now_secs() - inner.window_start).max(0.0) as u64;
+
+            let by_method: HashMap<String, Value> = inner
+                .by_method
+                .iter()
+                .map(|(k, v)| (k.clone(), v.to_json()))
+                .collect();
+            let by_endpoint: HashMap<String, Value> = inner
+                .by_endpoint
+                .iter()
+                .map(|(k, v)| (k.clone(), v.to_json()))
+                .collect();
+
+            let snap = MetricsSnapshot {
+                timestamp: Utc::now(),
+                window_seconds,
+                by_method,
+                by_endpoint,
+                by_status_class: inner.by_status_class.clone(),
+                error_codes: inner.error_codes.clone(),
+                totals: inner.totals.to_json(),
+            };
+
+            inner.snapshots.push(snap.clone());
+            let cutoff = now_secs() - (self.config.retention_hours * 3600) as f64;
+            inner
+                .snapshots
+                .retain(|s| s.timestamp.timestamp() as f64 > cutoff);
+
+            inner.by_method.clear();
+            inner.by_endpoint.clear();
+            inner.by_status_class.clear();
+            inner.error_codes.clear();
+            inner.totals = OperationStats::default();
+            inner.window_start = now_secs();
+
+            snap
+        };
+        let _ = snapshot;
+        self.save_snapshots();
+    }
+
+    fn save_snapshots(&self) {
+        let snapshots = { self.inner.lock().snapshots.clone() };
+        if let Some(parent) = self.snapshots_path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        let data = json!({ "snapshots": snapshots });
+        let _ = std::fs::write(
+            &self.snapshots_path,
+            serde_json::to_string_pretty(&data).unwrap_or_default(),
+        );
+    }
+
+    pub fn start_background(self: Arc<Self>) -> tokio::task::JoinHandle<()> {
+        let interval = std::time::Duration::from_secs(self.config.interval_minutes * 60);
+        tokio::spawn(async move {
+            let mut timer = tokio::time::interval(interval);
+            timer.tick().await;
+            loop {
+                timer.tick().await;
+                self.take_snapshot();
+            }
+        })
+    }
+}
+
+pub fn classify_endpoint(path: &str) -> &'static str {
+    if path.is_empty() || path == "/" {
+        return "service";
+    }
+    let trimmed = path.trim_end_matches('/');
+    if trimmed.starts_with("/ui") {
+        return "ui";
+    }
+    if trimmed.starts_with("/kms") {
+        return "kms";
+    }
+    if trimmed.starts_with("/myfsio") {
+        return "service";
+    }
+    let parts: Vec<&str> = trimmed.trim_start_matches('/').split('/').collect();
+    match parts.len() {
+        0 => "service",
+        1 => "bucket",
+        _ => "object",
+    }
+}
+
+fn now_secs() -> f64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .map(|d| d.as_secs_f64())
+        .unwrap_or(0.0)
+}

rust/myfsio-engine/crates/myfsio-server/src/services/mod.rs

@@ -0,0 +1,9 @@
+pub mod gc;
+pub mod lifecycle;
+pub mod integrity;
+pub mod metrics;
+pub mod replication;
+pub mod s3_client;
+pub mod site_registry;
+pub mod site_sync;
+pub mod website_domains;

rust/myfsio-engine/crates/myfsio-server/src/services/replication.rs

@@ -0,0 +1,604 @@
+use std::collections::HashMap;
+use std::path::{Path, PathBuf};
+use std::sync::Arc;
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
+
+use aws_sdk_s3::primitives::ByteStream;
+use parking_lot::Mutex;
+use serde::{Deserialize, Serialize};
+use tokio::sync::Semaphore;
+
+use myfsio_storage::fs_backend::FsStorageBackend;
+use myfsio_storage::traits::StorageEngine;
+
+use crate::services::s3_client::{build_client, check_endpoint_health, ClientOptions};
+use crate::stores::connections::{ConnectionStore, RemoteConnection};
+
+pub const MODE_NEW_ONLY: &str = "new_only";
+pub const MODE_ALL: &str = "all";
+pub const MODE_BIDIRECTIONAL: &str = "bidirectional";
+
+#[derive(Debug, Clone, Default, Serialize, Deserialize)]
+pub struct ReplicationStats {
+    #[serde(default)]
+    pub objects_synced: u64,
+    #[serde(default)]
+    pub objects_pending: u64,
+    #[serde(default)]
+    pub objects_orphaned: u64,
+    #[serde(default)]
+    pub bytes_synced: u64,
+    #[serde(default)]
+    pub last_sync_at: Option<f64>,
+    #[serde(default)]
+    pub last_sync_key: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ReplicationRule {
+    pub bucket_name: String,
+    pub target_connection_id: String,
+    pub target_bucket: String,
+    #[serde(default = "default_true")]
+    pub enabled: bool,
+    #[serde(default = "default_mode")]
+    pub mode: String,
+    #[serde(default)]
+    pub created_at: Option<f64>,
+    #[serde(default)]
+    pub stats: ReplicationStats,
+    #[serde(default = "default_true")]
+    pub sync_deletions: bool,
+    #[serde(default)]
+    pub last_pull_at: Option<f64>,
+    #[serde(default)]
+    pub filter_prefix: Option<String>,
+}
+
+fn default_true() -> bool {
+    true
+}
+fn default_mode() -> String {
+    MODE_NEW_ONLY.to_string()
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ReplicationFailure {
+    pub object_key: String,
+    pub error_message: String,
+    pub timestamp: f64,
+    pub failure_count: u32,
+    pub bucket_name: String,
+    pub action: String,
+    #[serde(default)]
+    pub last_error_code: Option<String>,
+}
+
+pub struct ReplicationFailureStore {
+    storage_root: PathBuf,
+    max_failures_per_bucket: usize,
+    cache: Mutex<HashMap<String, Vec<ReplicationFailure>>>,
+}
+
+impl ReplicationFailureStore {
+    pub fn new(storage_root: PathBuf, max_failures_per_bucket: usize) -> Self {
+        Self {
+            storage_root,
+            max_failures_per_bucket,
+            cache: Mutex::new(HashMap::new()),
+        }
+    }
+
+    fn path(&self, bucket: &str) -> PathBuf {
+        self.storage_root
+            .join(".myfsio.sys")
+            .join("buckets")
+            .join(bucket)
+            .join("replication_failures.json")
+    }
+
+    fn load_from_disk(&self, bucket: &str) -> Vec<ReplicationFailure> {
+        let path = self.path(bucket);
+        if !path.exists() {
+            return Vec::new();
+        }
+        match std::fs::read_to_string(&path) {
+            Ok(text) => {
+                let parsed: serde_json::Value = match serde_json::from_str(&text) {
+                    Ok(v) => v,
+                    Err(_) => return Vec::new(),
+                };
+                parsed
+                    .get("failures")
+                    .and_then(|v| serde_json::from_value(v.clone()).ok())
+                    .unwrap_or_default()
+            }
+            Err(_) => Vec::new(),
+        }
+    }
+
+    fn save_to_disk(&self, bucket: &str, failures: &[ReplicationFailure]) {
+        let path = self.path(bucket);
+        if let Some(parent) = path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        let trimmed = &failures[..failures.len().min(self.max_failures_per_bucket)];
+        let data = serde_json::json!({ "failures": trimmed });
+        let _ = std::fs::write(&path, serde_json::to_string_pretty(&data).unwrap_or_default());
+    }
+
+    pub fn load(&self, bucket: &str) -> Vec<ReplicationFailure> {
+        let mut cache = self.cache.lock();
+        if let Some(existing) = cache.get(bucket) {
+            return existing.clone();
+        }
+        let loaded = self.load_from_disk(bucket);
+        cache.insert(bucket.to_string(), loaded.clone());
+        loaded
+    }
+
+    pub fn save(&self, bucket: &str, failures: Vec<ReplicationFailure>) {
+        let trimmed: Vec<ReplicationFailure> = failures
+            .into_iter()
+            .take(self.max_failures_per_bucket)
+            .collect();
+        self.save_to_disk(bucket, &trimmed);
+        self.cache.lock().insert(bucket.to_string(), trimmed);
+    }
+
+    pub fn add(&self, bucket: &str, failure: ReplicationFailure) {
+        let mut failures = self.load(bucket);
+        if let Some(existing) = failures.iter_mut().find(|f| f.object_key == failure.object_key) {
+            existing.failure_count += 1;
+            existing.timestamp = failure.timestamp;
+            existing.error_message = failure.error_message.clone();
+            existing.last_error_code = failure.last_error_code.clone();
+        } else {
+            failures.insert(0, failure);
+        }
+        self.save(bucket, failures);
+    }
+
+    pub fn remove(&self, bucket: &str, object_key: &str) -> bool {
+        let failures = self.load(bucket);
+        let before = failures.len();
+        let after: Vec<_> = failures
+            .into_iter()
+            .filter(|f| f.object_key != object_key)
+            .collect();
+        if after.len() != before {
+            self.save(bucket, after);
+            true
+        } else {
+            false
+        }
+    }
+
+    pub fn clear(&self, bucket: &str) {
+        self.cache.lock().remove(bucket);
+        let path = self.path(bucket);
+        let _ = std::fs::remove_file(path);
+    }
+
+    pub fn get(&self, bucket: &str, object_key: &str) -> Option<ReplicationFailure> {
+        self.load(bucket)
+            .into_iter()
+            .find(|f| f.object_key == object_key)
+    }
+
+    pub fn count(&self, bucket: &str) -> usize {
+        self.load(bucket).len()
+    }
+}
+
+pub struct ReplicationManager {
+    storage: Arc<FsStorageBackend>,
+    connections: Arc<ConnectionStore>,
+    rules_path: PathBuf,
+    rules: Mutex<HashMap<String, ReplicationRule>>,
+    client_options: ClientOptions,
+    streaming_threshold_bytes: u64,
+    pub failures: Arc<ReplicationFailureStore>,
+    semaphore: Arc<Semaphore>,
+}
+
+impl ReplicationManager {
+    pub fn new(
+        storage: Arc<FsStorageBackend>,
+        connections: Arc<ConnectionStore>,
+        storage_root: &Path,
+        connect_timeout: Duration,
+        read_timeout: Duration,
+        max_retries: u32,
+        streaming_threshold_bytes: u64,
+        max_failures_per_bucket: usize,
+    ) -> Self {
+        let rules_path = storage_root
+            .join(".myfsio.sys")
+            .join("config")
+            .join("replication_rules.json");
+        let rules = load_rules(&rules_path);
+        let failures = Arc::new(ReplicationFailureStore::new(
+            storage_root.to_path_buf(),
+            max_failures_per_bucket,
+        ));
+        let client_options = ClientOptions {
+            connect_timeout,
+            read_timeout,
+            max_attempts: max_retries,
+        };
+        Self {
+            storage,
+            connections,
+            rules_path,
+            rules: Mutex::new(rules),
+            client_options,
+            streaming_threshold_bytes,
+            failures,
+            semaphore: Arc::new(Semaphore::new(4)),
+        }
+    }
+
+    pub fn reload_rules(&self) {
+        *self.rules.lock() = load_rules(&self.rules_path);
+    }
+
+    pub fn list_rules(&self) -> Vec<ReplicationRule> {
+        self.rules.lock().values().cloned().collect()
+    }
+
+    pub fn get_rule(&self, bucket: &str) -> Option<ReplicationRule> {
+        self.rules.lock().get(bucket).cloned()
+    }
+
+    pub fn set_rule(&self, rule: ReplicationRule) {
+        {
+            let mut guard = self.rules.lock();
+            guard.insert(rule.bucket_name.clone(), rule);
+        }
+        self.save_rules();
+    }
+
+    pub fn delete_rule(&self, bucket: &str) {
+        {
+            let mut guard = self.rules.lock();
+            guard.remove(bucket);
+        }
+        self.save_rules();
+    }
+
+    pub fn save_rules(&self) {
+        let snapshot: HashMap<String, ReplicationRule> = self.rules.lock().clone();
+        if let Some(parent) = self.rules_path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        if let Ok(text) = serde_json::to_string_pretty(&snapshot) {
+            let _ = std::fs::write(&self.rules_path, text);
+        }
+    }
+
+    fn update_last_sync(&self, bucket: &str, key: &str) {
+        {
+            let mut guard = self.rules.lock();
+            if let Some(rule) = guard.get_mut(bucket) {
+                rule.stats.last_sync_at = Some(now_secs());
+                rule.stats.last_sync_key = Some(key.to_string());
+            }
+        }
+        self.save_rules();
+    }
+
+    pub async fn trigger(self: Arc<Self>, bucket: String, key: String, action: String) {
+        let rule = match self.get_rule(&bucket) {
+            Some(r) if r.enabled => r,
+            _ => return,
+        };
+        let connection = match self.connections.get(&rule.target_connection_id) {
+            Some(c) => c,
+            None => {
+                tracing::warn!(
+                    "Replication skipped for {}/{}: connection {} not found",
+                    bucket,
+                    key,
+                    rule.target_connection_id
+                );
+                return;
+            }
+        };
+        let permit = match self.semaphore.clone().try_acquire_owned() {
+            Ok(p) => p,
+            Err(_) => {
+                let sem = self.semaphore.clone();
+                match sem.acquire_owned().await {
+                    Ok(p) => p,
+                    Err(_) => return,
+                }
+            }
+        };
+        let manager = self.clone();
+        tokio::spawn(async move {
+            let _permit = permit;
+            manager.replicate_task(&bucket, &key, &rule, &connection, &action).await;
+        });
+    }
+
+    async fn replicate_task(
+        &self,
+        bucket: &str,
+        object_key: &str,
+        rule: &ReplicationRule,
+        conn: &RemoteConnection,
+        action: &str,
+    ) {
+        if object_key.contains("..") || object_key.starts_with('/') || object_key.starts_with('\\') {
+            tracing::error!("Invalid object key (path traversal): {}", object_key);
+            return;
+        }
+
+        let client = build_client(conn, &self.client_options);
+
+        if action == "delete" {
+            match client
+                .delete_object()
+                .bucket(&rule.target_bucket)
+                .key(object_key)
+                .send()
+                .await
+            {
+                Ok(_) => {
+                    tracing::info!(
+                        "Replicated DELETE {}/{} to {} ({})",
+                        bucket,
+                        object_key,
+                        conn.name,
+                        rule.target_bucket
+                    );
+                    self.update_last_sync(bucket, object_key);
+                    self.failures.remove(bucket, object_key);
+                }
+                Err(err) => {
+                    let msg = format!("{:?}", err);
+                    tracing::error!("Replication DELETE failed {}/{}: {}", bucket, object_key, msg);
+                    self.failures.add(
+                        bucket,
+                        ReplicationFailure {
+                            object_key: object_key.to_string(),
+                            error_message: msg,
+                            timestamp: now_secs(),
+                            failure_count: 1,
+                            bucket_name: bucket.to_string(),
+                            action: "delete".to_string(),
+                            last_error_code: None,
+                        },
+                    );
+                }
+            }
+            return;
+        }
+
+        let src_path = match self.storage.get_object_path(bucket, object_key).await {
+            Ok(p) => p,
+            Err(_) => {
+                tracing::error!("Source object not found: {}/{}", bucket, object_key);
+                return;
+            }
+        };
+        let file_size = match tokio::fs::metadata(&src_path).await {
+            Ok(m) => m.len(),
+            Err(_) => 0,
+        };
+        let content_type = mime_guess::from_path(&src_path)
+            .first_raw()
+            .map(|s| s.to_string());
+
+        let upload_result = upload_object(
+            &client,
+            &rule.target_bucket,
+            object_key,
+            &src_path,
+            file_size,
+            self.streaming_threshold_bytes,
+            content_type.as_deref(),
+        )
+        .await;
+
+        let final_result = match upload_result {
+            Err(err) if is_no_such_bucket(&err) => {
+                tracing::info!(
+                    "Target bucket {} not found, creating it",
+                    rule.target_bucket
+                );
+                match client
+                    .create_bucket()
+                    .bucket(&rule.target_bucket)
+                    .send()
+                    .await
+                {
+                    Ok(_) | Err(_) => upload_object(
+                        &client,
+                        &rule.target_bucket,
+                        object_key,
+                        &src_path,
+                        file_size,
+                        self.streaming_threshold_bytes,
+                        content_type.as_deref(),
+                    )
+                    .await,
+                }
+            }
+            other => other,
+        };
+
+        match final_result {
+            Ok(()) => {
+                tracing::info!(
+                    "Replicated {}/{} to {} ({})",
+                    bucket,
+                    object_key,
+                    conn.name,
+                    rule.target_bucket
+                );
+                self.update_last_sync(bucket, object_key);
+                self.failures.remove(bucket, object_key);
+            }
+            Err(err) => {
+                let msg = err.to_string();
+                tracing::error!("Replication failed {}/{}: {}", bucket, object_key, msg);
+                self.failures.add(
+                    bucket,
+                    ReplicationFailure {
+                        object_key: object_key.to_string(),
+                        error_message: msg,
+                        timestamp: now_secs(),
+                        failure_count: 1,
+                        bucket_name: bucket.to_string(),
+                        action: action.to_string(),
+                        last_error_code: None,
+                    },
+                );
+            }
+        }
+    }
+
+    pub async fn check_endpoint(&self, conn: &RemoteConnection) -> bool {
+        let client = build_client(conn, &self.client_options);
+        check_endpoint_health(&client).await
+    }
+
+    pub async fn retry_failed(&self, bucket: &str, object_key: &str) -> bool {
+        let failure = match self.failures.get(bucket, object_key) {
+            Some(f) => f,
+            None => return false,
+        };
+        let rule = match self.get_rule(bucket) {
+            Some(r) if r.enabled => r,
+            _ => return false,
+        };
+        let conn = match self.connections.get(&rule.target_connection_id) {
+            Some(c) => c,
+            None => return false,
+        };
+        self.replicate_task(bucket, object_key, &rule, &conn, &failure.action)
+            .await;
+        true
+    }
+
+    pub async fn retry_all(&self, bucket: &str) -> (usize, usize) {
+        let failures = self.failures.load(bucket);
+        if failures.is_empty() {
+            return (0, 0);
+        }
+        let rule = match self.get_rule(bucket) {
+            Some(r) if r.enabled => r,
+            _ => return (0, failures.len()),
+        };
+        let conn = match self.connections.get(&rule.target_connection_id) {
+            Some(c) => c,
+            None => return (0, failures.len()),
+        };
+        let mut submitted = 0;
+        for failure in failures {
+            self.replicate_task(bucket, &failure.object_key, &rule, &conn, &failure.action)
+                .await;
+            submitted += 1;
+        }
+        (submitted, 0)
+    }
+
+    pub fn get_failure_count(&self, bucket: &str) -> usize {
+        self.failures.count(bucket)
+    }
+
+    pub fn get_failed_items(
+        &self,
+        bucket: &str,
+        limit: usize,
+        offset: usize,
+    ) -> Vec<ReplicationFailure> {
+        self.failures
+            .load(bucket)
+            .into_iter()
+            .skip(offset)
+            .take(limit)
+            .collect()
+    }
+
+    pub fn dismiss_failure(&self, bucket: &str, key: &str) -> bool {
+        self.failures.remove(bucket, key)
+    }
+
+    pub fn clear_failures(&self, bucket: &str) {
+        self.failures.clear(bucket);
+    }
+
+    pub fn rules_snapshot(&self) -> HashMap<String, ReplicationRule> {
+        self.rules.lock().clone()
+    }
+
+    pub fn update_last_pull(&self, bucket: &str, at: f64) {
+        {
+            let mut guard = self.rules.lock();
+            if let Some(rule) = guard.get_mut(bucket) {
+                rule.last_pull_at = Some(at);
+            }
+        }
+        self.save_rules();
+    }
+
+    pub fn client_options(&self) -> &ClientOptions {
+        &self.client_options
+    }
+}
+
+fn is_no_such_bucket<E: std::fmt::Debug>(err: &E) -> bool {
+    let text = format!("{:?}", err);
+    text.contains("NoSuchBucket")
+}
+
+async fn upload_object(
+    client: &aws_sdk_s3::Client,
+    bucket: &str,
+    key: &str,
+    path: &Path,
+    file_size: u64,
+    streaming_threshold: u64,
+    content_type: Option<&str>,
+) -> Result<(), aws_sdk_s3::error::SdkError<aws_sdk_s3::operation::put_object::PutObjectError>> {
+    let mut req = client.put_object().bucket(bucket).key(key);
+    if let Some(ct) = content_type {
+        req = req.content_type(ct);
+    }
+
+    let body = if file_size >= streaming_threshold {
+        ByteStream::from_path(path).await.map_err(|e| {
+            aws_sdk_s3::error::SdkError::construction_failure(Box::new(std::io::Error::new(
+                std::io::ErrorKind::Other,
+                e,
+            )))
+        })?
+    } else {
+        let bytes = tokio::fs::read(path).await.map_err(|e| {
+            aws_sdk_s3::error::SdkError::construction_failure(Box::new(e))
+        })?;
+        ByteStream::from(bytes)
+    };
+
+    req.body(body).send().await.map(|_| ())
+}
+
+fn load_rules(path: &Path) -> HashMap<String, ReplicationRule> {
+    if !path.exists() {
+        return HashMap::new();
+    }
+    match std::fs::read_to_string(path) {
+        Ok(text) => serde_json::from_str(&text).unwrap_or_default(),
+        Err(_) => HashMap::new(),
+    }
+}
+
+fn now_secs() -> f64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .map(|d| d.as_secs_f64())
+        .unwrap_or(0.0)
+}

rust/myfsio-engine/crates/myfsio-server/src/services/s3_client.rs

@@ -0,0 +1,64 @@
+use std::time::Duration;
+
+use aws_config::BehaviorVersion;
+use aws_credential_types::Credentials;
+use aws_sdk_s3::config::{Region, SharedCredentialsProvider};
+use aws_sdk_s3::Client;
+
+use crate::stores::connections::RemoteConnection;
+
+pub struct ClientOptions {
+    pub connect_timeout: Duration,
+    pub read_timeout: Duration,
+    pub max_attempts: u32,
+}
+
+impl Default for ClientOptions {
+    fn default() -> Self {
+        Self {
+            connect_timeout: Duration::from_secs(5),
+            read_timeout: Duration::from_secs(30),
+            max_attempts: 2,
+        }
+    }
+}
+
+pub fn build_client(connection: &RemoteConnection, options: &ClientOptions) -> Client {
+    let credentials = Credentials::new(
+        connection.access_key.clone(),
+        connection.secret_key.clone(),
+        None,
+        None,
+        "myfsio-replication",
+    );
+
+    let timeout_config = aws_smithy_types::timeout::TimeoutConfig::builder()
+        .connect_timeout(options.connect_timeout)
+        .read_timeout(options.read_timeout)
+        .build();
+
+    let retry_config = aws_smithy_types::retry::RetryConfig::standard()
+        .with_max_attempts(options.max_attempts);
+
+    let config = aws_sdk_s3::config::Builder::new()
+        .behavior_version(BehaviorVersion::latest())
+        .credentials_provider(SharedCredentialsProvider::new(credentials))
+        .region(Region::new(connection.region.clone()))
+        .endpoint_url(connection.endpoint_url.clone())
+        .force_path_style(true)
+        .timeout_config(timeout_config)
+        .retry_config(retry_config)
+        .build();
+
+    Client::from_conf(config)
+}
+
+pub async fn check_endpoint_health(client: &Client) -> bool {
+    match client.list_buckets().send().await {
+        Ok(_) => true,
+        Err(err) => {
+            tracing::warn!("Endpoint health check failed: {:?}", err);
+            false
+        }
+    }
+}

rust/myfsio-engine/crates/myfsio-server/src/services/site_registry.rs

@@ -0,0 +1,143 @@
+use chrono::Utc;
+use parking_lot::RwLock;
+use serde::{Deserialize, Serialize};
+use std::path::PathBuf;
+use std::sync::Arc;
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct SiteInfo {
+    pub site_id: String,
+    pub endpoint: String,
+    #[serde(default = "default_region")]
+    pub region: String,
+    #[serde(default = "default_priority")]
+    pub priority: i32,
+    #[serde(default)]
+    pub display_name: String,
+    #[serde(default)]
+    pub created_at: Option<String>,
+}
+
+fn default_region() -> String {
+    "us-east-1".to_string()
+}
+fn default_priority() -> i32 {
+    100
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct PeerSite {
+    pub site_id: String,
+    pub endpoint: String,
+    #[serde(default = "default_region")]
+    pub region: String,
+    #[serde(default = "default_priority")]
+    pub priority: i32,
+    #[serde(default)]
+    pub display_name: String,
+    #[serde(default)]
+    pub connection_id: Option<String>,
+    #[serde(default)]
+    pub created_at: Option<String>,
+    #[serde(default)]
+    pub is_healthy: bool,
+    #[serde(default)]
+    pub last_health_check: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+struct RegistryData {
+    #[serde(default)]
+    local: Option<SiteInfo>,
+    #[serde(default)]
+    peers: Vec<PeerSite>,
+}
+
+pub struct SiteRegistry {
+    path: PathBuf,
+    data: Arc<RwLock<RegistryData>>,
+}
+
+impl SiteRegistry {
+    pub fn new(storage_root: &std::path::Path) -> Self {
+        let path = storage_root
+            .join(".myfsio.sys")
+            .join("config")
+            .join("site_registry.json");
+        let data = if path.exists() {
+            std::fs::read_to_string(&path)
+                .ok()
+                .and_then(|s| serde_json::from_str(&s).ok())
+                .unwrap_or_default()
+        } else {
+            RegistryData::default()
+        };
+        Self {
+            path,
+            data: Arc::new(RwLock::new(data)),
+        }
+    }
+
+    fn save(&self) {
+        let data = self.data.read();
+        if let Some(parent) = self.path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        if let Ok(json) = serde_json::to_string_pretty(&*data) {
+            let _ = std::fs::write(&self.path, json);
+        }
+    }
+
+    pub fn get_local_site(&self) -> Option<SiteInfo> {
+        self.data.read().local.clone()
+    }
+
+    pub fn set_local_site(&self, site: SiteInfo) {
+        self.data.write().local = Some(site);
+        self.save();
+    }
+
+    pub fn list_peers(&self) -> Vec<PeerSite> {
+        self.data.read().peers.clone()
+    }
+
+    pub fn get_peer(&self, site_id: &str) -> Option<PeerSite> {
+        self.data.read().peers.iter().find(|p| p.site_id == site_id).cloned()
+    }
+
+    pub fn add_peer(&self, peer: PeerSite) {
+        self.data.write().peers.push(peer);
+        self.save();
+    }
+
+    pub fn update_peer(&self, peer: PeerSite) {
+        let mut data = self.data.write();
+        if let Some(existing) = data.peers.iter_mut().find(|p| p.site_id == peer.site_id) {
+            *existing = peer;
+        }
+        drop(data);
+        self.save();
+    }
+
+    pub fn delete_peer(&self, site_id: &str) -> bool {
+        let mut data = self.data.write();
+        let len_before = data.peers.len();
+        data.peers.retain(|p| p.site_id != site_id);
+        let removed = data.peers.len() < len_before;
+        drop(data);
+        if removed {
+            self.save();
+        }
+        removed
+    }
+
+    pub fn update_health(&self, site_id: &str, is_healthy: bool) {
+        let mut data = self.data.write();
+        if let Some(peer) = data.peers.iter_mut().find(|p| p.site_id == site_id) {
+            peer.is_healthy = is_healthy;
+            peer.last_health_check = Some(Utc::now().to_rfc3339());
+        }
+        drop(data);
+        self.save();
+    }
+}

rust/myfsio-engine/crates/myfsio-server/src/services/site_sync.rs

@@ -0,0 +1,485 @@
+use std::collections::HashMap;
+use std::path::PathBuf;
+use std::pin::Pin;
+use std::sync::Arc;
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
+
+use aws_sdk_s3::Client;
+use parking_lot::Mutex;
+use serde::{Deserialize, Serialize};
+use tokio::io::AsyncRead;
+use tokio::sync::Notify;
+
+use myfsio_common::types::{ListParams, ObjectMeta};
+use myfsio_storage::fs_backend::FsStorageBackend;
+use myfsio_storage::traits::StorageEngine;
+
+use crate::services::replication::{ReplicationManager, ReplicationRule, MODE_BIDIRECTIONAL};
+use crate::services::s3_client::{build_client, ClientOptions};
+use crate::stores::connections::ConnectionStore;
+
+#[derive(Debug, Clone, Default, Serialize, Deserialize)]
+pub struct SyncedObjectInfo {
+    pub last_synced_at: f64,
+    pub remote_etag: String,
+    pub source: String,
+}
+
+#[derive(Debug, Clone, Default, Serialize, Deserialize)]
+pub struct SyncState {
+    #[serde(default)]
+    pub synced_objects: HashMap<String, SyncedObjectInfo>,
+    #[serde(default)]
+    pub last_full_sync: Option<f64>,
+}
+
+#[derive(Debug, Clone, Default, Serialize)]
+pub struct SiteSyncStats {
+    pub last_sync_at: Option<f64>,
+    pub objects_pulled: u64,
+    pub objects_skipped: u64,
+    pub conflicts_resolved: u64,
+    pub deletions_applied: u64,
+    pub errors: u64,
+}
+
+#[derive(Debug, Clone)]
+struct RemoteObjectMeta {
+    last_modified: f64,
+    etag: String,
+}
+
+pub struct SiteSyncWorker {
+    storage: Arc<FsStorageBackend>,
+    connections: Arc<ConnectionStore>,
+    replication: Arc<ReplicationManager>,
+    storage_root: PathBuf,
+    interval: Duration,
+    batch_size: usize,
+    clock_skew_tolerance: f64,
+    client_options: ClientOptions,
+    bucket_stats: Mutex<HashMap<String, SiteSyncStats>>,
+    shutdown: Arc<Notify>,
+}
+
+impl SiteSyncWorker {
+    pub fn new(
+        storage: Arc<FsStorageBackend>,
+        connections: Arc<ConnectionStore>,
+        replication: Arc<ReplicationManager>,
+        storage_root: PathBuf,
+        interval_seconds: u64,
+        batch_size: usize,
+        connect_timeout: Duration,
+        read_timeout: Duration,
+        max_retries: u32,
+        clock_skew_tolerance: f64,
+    ) -> Self {
+        Self {
+            storage,
+            connections,
+            replication,
+            storage_root,
+            interval: Duration::from_secs(interval_seconds),
+            batch_size,
+            clock_skew_tolerance,
+            client_options: ClientOptions {
+                connect_timeout,
+                read_timeout,
+                max_attempts: max_retries,
+            },
+            bucket_stats: Mutex::new(HashMap::new()),
+            shutdown: Arc::new(Notify::new()),
+        }
+    }
+
+    pub fn shutdown(&self) {
+        self.shutdown.notify_waiters();
+    }
+
+    pub fn get_stats(&self, bucket: &str) -> Option<SiteSyncStats> {
+        self.bucket_stats.lock().get(bucket).cloned()
+    }
+
+    pub async fn run(self: Arc<Self>) {
+        tracing::info!("Site sync worker started (interval={}s)", self.interval.as_secs());
+        loop {
+            tokio::select! {
+                _ = tokio::time::sleep(self.interval) => {}
+                _ = self.shutdown.notified() => {
+                    tracing::info!("Site sync worker shutting down");
+                    return;
+                }
+            }
+            self.run_cycle().await;
+        }
+    }
+
+    async fn run_cycle(&self) {
+        let rules = self.replication.rules_snapshot();
+        for (bucket, rule) in rules {
+            if rule.mode != MODE_BIDIRECTIONAL || !rule.enabled {
+                continue;
+            }
+            match self.sync_bucket(&rule).await {
+                Ok(stats) => {
+                    self.bucket_stats.lock().insert(bucket, stats);
+                }
+                Err(e) => {
+                    tracing::error!("Site sync failed for bucket {}: {}", bucket, e);
+                }
+            }
+        }
+    }
+
+    pub async fn trigger_sync(&self, bucket: &str) -> Option<SiteSyncStats> {
+        let rule = self.replication.get_rule(bucket)?;
+        if rule.mode != MODE_BIDIRECTIONAL || !rule.enabled {
+            return None;
+        }
+        match self.sync_bucket(&rule).await {
+            Ok(stats) => {
+                self.bucket_stats
+                    .lock()
+                    .insert(bucket.to_string(), stats.clone());
+                Some(stats)
+            }
+            Err(e) => {
+                tracing::error!("Site sync trigger failed for {}: {}", bucket, e);
+                None
+            }
+        }
+    }
+
+    async fn sync_bucket(&self, rule: &ReplicationRule) -> Result<SiteSyncStats, String> {
+        let mut stats = SiteSyncStats::default();
+        let connection = self
+            .connections
+            .get(&rule.target_connection_id)
+            .ok_or_else(|| format!("connection {} not found", rule.target_connection_id))?;
+
+        let local_objects = self
+            .list_local_objects(&rule.bucket_name)
+            .await
+            .map_err(|e| format!("list local failed: {}", e))?;
+
+        let client = build_client(&connection, &self.client_options);
+        let remote_objects = self
+            .list_remote_objects(&client, &rule.target_bucket)
+            .await
+            .map_err(|e| format!("list remote failed: {}", e))?;
+
+        let mut sync_state = self.load_sync_state(&rule.bucket_name);
+
+        let mut to_pull: Vec<String> = Vec::new();
+        for (key, remote_meta) in &remote_objects {
+            if let Some(local_meta) = local_objects.get(key) {
+                match self.resolve_conflict(local_meta, remote_meta) {
+                    "pull" => {
+                        to_pull.push(key.clone());
+                        stats.conflicts_resolved += 1;
+                    }
+                    _ => {
+                        stats.objects_skipped += 1;
+                    }
+                }
+            } else {
+                to_pull.push(key.clone());
+            }
+        }
+
+        let mut pulled = 0usize;
+        for key in &to_pull {
+            if pulled >= self.batch_size {
+                break;
+            }
+            let remote_meta = match remote_objects.get(key) {
+                Some(m) => m,
+                None => continue,
+            };
+            if self
+                .pull_object(&client, &rule.target_bucket, &rule.bucket_name, key)
+                .await
+            {
+                stats.objects_pulled += 1;
+                pulled += 1;
+                sync_state.synced_objects.insert(
+                    key.clone(),
+                    SyncedObjectInfo {
+                        last_synced_at: now_secs(),
+                        remote_etag: remote_meta.etag.clone(),
+                        source: "remote".to_string(),
+                    },
+                );
+            } else {
+                stats.errors += 1;
+            }
+        }
+
+        if rule.sync_deletions {
+            let tracked_keys: Vec<String> = sync_state.synced_objects.keys().cloned().collect();
+            for key in tracked_keys {
+                if remote_objects.contains_key(&key) {
+                    continue;
+                }
+                let local_meta = match local_objects.get(&key) {
+                    Some(m) => m,
+                    None => continue,
+                };
+                let tracked = match sync_state.synced_objects.get(&key) {
+                    Some(t) => t.clone(),
+                    None => continue,
+                };
+                if tracked.source != "remote" {
+                    continue;
+                }
+                let local_ts = local_meta.last_modified.timestamp() as f64;
+                if local_ts <= tracked.last_synced_at
+                    && self.apply_remote_deletion(&rule.bucket_name, &key).await
+                {
+                    stats.deletions_applied += 1;
+                    sync_state.synced_objects.remove(&key);
+                }
+            }
+        }
+
+        sync_state.last_full_sync = Some(now_secs());
+        self.save_sync_state(&rule.bucket_name, &sync_state);
+
+        self.replication
+            .update_last_pull(&rule.bucket_name, now_secs());
+
+        stats.last_sync_at = Some(now_secs());
+        tracing::info!(
+            "Site sync completed for {}: pulled={}, skipped={}, conflicts={}, deletions={}, errors={}",
+            rule.bucket_name,
+            stats.objects_pulled,
+            stats.objects_skipped,
+            stats.conflicts_resolved,
+            stats.deletions_applied,
+            stats.errors,
+        );
+        Ok(stats)
+    }
+
+    async fn list_local_objects(
+        &self,
+        bucket: &str,
+    ) -> Result<HashMap<String, ObjectMeta>, String> {
+        let mut result = HashMap::new();
+        let mut token: Option<String> = None;
+        loop {
+            let params = ListParams {
+                max_keys: 1000,
+                continuation_token: token.clone(),
+                prefix: None,
+                start_after: None,
+            };
+            let page = self
+                .storage
+                .list_objects(bucket, &params)
+                .await
+                .map_err(|e| e.to_string())?;
+            for obj in page.objects {
+                result.insert(obj.key.clone(), obj);
+            }
+            if !page.is_truncated {
+                break;
+            }
+            token = page.next_continuation_token;
+            if token.is_none() {
+                break;
+            }
+        }
+        Ok(result)
+    }
+
+    async fn list_remote_objects(
+        &self,
+        client: &Client,
+        bucket: &str,
+    ) -> Result<HashMap<String, RemoteObjectMeta>, String> {
+        let mut result = HashMap::new();
+        let mut continuation: Option<String> = None;
+        loop {
+            let mut req = client.list_objects_v2().bucket(bucket);
+            if let Some(ref t) = continuation {
+                req = req.continuation_token(t);
+            }
+            let resp = match req.send().await {
+                Ok(r) => r,
+                Err(err) => {
+                    let msg = format!("{:?}", err);
+                    if msg.contains("NoSuchBucket") {
+                        return Ok(result);
+                    }
+                    return Err(msg);
+                }
+            };
+            for obj in resp.contents() {
+                let key = match obj.key() {
+                    Some(k) => k.to_string(),
+                    None => continue,
+                };
+                let last_modified = obj
+                    .last_modified()
+                    .and_then(|t| {
+                        let secs = t.secs();
+                        let nanos = t.subsec_nanos();
+                        Some(secs as f64 + nanos as f64 / 1_000_000_000.0)
+                    })
+                    .unwrap_or(0.0);
+                let etag = obj.e_tag().unwrap_or("").trim_matches('"').to_string();
+                result.insert(
+                    key,
+                    RemoteObjectMeta {
+                        last_modified,
+                        etag,
+                    },
+                );
+            }
+            if resp.is_truncated().unwrap_or(false) {
+                continuation = resp.next_continuation_token().map(|s| s.to_string());
+                if continuation.is_none() {
+                    break;
+                }
+            } else {
+                break;
+            }
+        }
+        Ok(result)
+    }
+
+    fn resolve_conflict(&self, local: &ObjectMeta, remote: &RemoteObjectMeta) -> &'static str {
+        let local_ts = local.last_modified.timestamp() as f64
+            + local.last_modified.timestamp_subsec_nanos() as f64 / 1_000_000_000.0;
+        let remote_ts = remote.last_modified;
+
+        if (remote_ts - local_ts).abs() < self.clock_skew_tolerance {
+            let local_etag = local.etag.clone().unwrap_or_default();
+            let local_etag_trim = local_etag.trim_matches('"');
+            if remote.etag == local_etag_trim {
+                return "skip";
+            }
+            if remote.etag.as_str() > local_etag_trim {
+                return "pull";
+            }
+            return "keep";
+        }
+
+        if remote_ts > local_ts {
+            "pull"
+        } else {
+            "keep"
+        }
+    }
+
+    async fn pull_object(
+        &self,
+        client: &Client,
+        remote_bucket: &str,
+        local_bucket: &str,
+        key: &str,
+    ) -> bool {
+        let resp = match client
+            .get_object()
+            .bucket(remote_bucket)
+            .key(key)
+            .send()
+            .await
+        {
+            Ok(r) => r,
+            Err(err) => {
+                tracing::error!("Pull GetObject failed {}/{}: {:?}", local_bucket, key, err);
+                return false;
+            }
+        };
+
+        let head = match client
+            .head_object()
+            .bucket(remote_bucket)
+            .key(key)
+            .send()
+            .await
+        {
+            Ok(r) => r,
+            Err(err) => {
+                tracing::error!("Pull HeadObject failed {}/{}: {:?}", local_bucket, key, err);
+                return false;
+            }
+        };
+
+        let metadata: Option<HashMap<String, String>> = head.metadata().map(|m| {
+            m.iter()
+                .map(|(k, v)| (k.clone(), v.clone()))
+                .collect()
+        });
+
+        let stream = resp.body.into_async_read();
+        let boxed: Pin<Box<dyn AsyncRead + Send>> = Box::pin(stream);
+
+        match self
+            .storage
+            .put_object(local_bucket, key, boxed, metadata)
+            .await
+        {
+            Ok(_) => {
+                tracing::debug!("Pulled object {}/{} from remote", local_bucket, key);
+                true
+            }
+            Err(err) => {
+                tracing::error!("Store pulled object failed {}/{}: {}", local_bucket, key, err);
+                false
+            }
+        }
+    }
+
+    async fn apply_remote_deletion(&self, bucket: &str, key: &str) -> bool {
+        match self.storage.delete_object(bucket, key).await {
+            Ok(_) => {
+                tracing::debug!("Applied remote deletion for {}/{}", bucket, key);
+                true
+            }
+            Err(err) => {
+                tracing::error!("Remote deletion failed {}/{}: {}", bucket, key, err);
+                false
+            }
+        }
+    }
+
+    fn sync_state_path(&self, bucket: &str) -> PathBuf {
+        self.storage_root
+            .join(".myfsio.sys")
+            .join("buckets")
+            .join(bucket)
+            .join("site_sync_state.json")
+    }
+
+    fn load_sync_state(&self, bucket: &str) -> SyncState {
+        let path = self.sync_state_path(bucket);
+        if !path.exists() {
+            return SyncState::default();
+        }
+        match std::fs::read_to_string(&path) {
+            Ok(text) => serde_json::from_str(&text).unwrap_or_default(),
+            Err(_) => SyncState::default(),
+        }
+    }
+
+    fn save_sync_state(&self, bucket: &str, state: &SyncState) {
+        let path = self.sync_state_path(bucket);
+        if let Some(parent) = path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        if let Ok(text) = serde_json::to_string_pretty(state) {
+            let _ = std::fs::write(&path, text);
+        }
+    }
+}
+
+fn now_secs() -> f64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .map(|d| d.as_secs_f64())
+        .unwrap_or(0.0)
+}

rust/myfsio-engine/crates/myfsio-server/src/services/website_domains.rs

@@ -0,0 +1,104 @@
+use parking_lot::RwLock;
+use serde::{Deserialize, Serialize};
+use std::collections::HashMap;
+use std::path::PathBuf;
+use std::sync::Arc;
+
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+struct DomainData {
+    #[serde(default)]
+    mappings: HashMap<String, String>,
+}
+
+pub struct WebsiteDomainStore {
+    path: PathBuf,
+    data: Arc<RwLock<DomainData>>,
+}
+
+impl WebsiteDomainStore {
+    pub fn new(storage_root: &std::path::Path) -> Self {
+        let path = storage_root
+            .join(".myfsio.sys")
+            .join("config")
+            .join("website_domains.json");
+        let data = if path.exists() {
+            std::fs::read_to_string(&path)
+                .ok()
+                .and_then(|s| serde_json::from_str(&s).ok())
+                .unwrap_or_default()
+        } else {
+            DomainData::default()
+        };
+        Self {
+            path,
+            data: Arc::new(RwLock::new(data)),
+        }
+    }
+
+    fn save(&self) {
+        let data = self.data.read();
+        if let Some(parent) = self.path.parent() {
+            let _ = std::fs::create_dir_all(parent);
+        }
+        if let Ok(json) = serde_json::to_string_pretty(&*data) {
+            let _ = std::fs::write(&self.path, json);
+        }
+    }
+
+    pub fn list_all(&self) -> Vec<serde_json::Value> {
+        self.data
+            .read()
+            .mappings
+            .iter()
+            .map(|(domain, bucket)| {
+                serde_json::json!({
+                    "domain": domain,
+                    "bucket": bucket,
+                })
+            })
+            .collect()
+    }
+
+    pub fn get_bucket(&self, domain: &str) -> Option<String> {
+        self.data.read().mappings.get(domain).cloned()
+    }
+
+    pub fn set_mapping(&self, domain: &str, bucket: &str) {
+        self.data.write().mappings.insert(domain.to_string(), bucket.to_string());
+        self.save();
+    }
+
+    pub fn delete_mapping(&self, domain: &str) -> bool {
+        let removed = self.data.write().mappings.remove(domain).is_some();
+        if removed {
+            self.save();
+        }
+        removed
+    }
+}
+
+pub fn normalize_domain(domain: &str) -> String {
+    domain.trim().to_ascii_lowercase()
+}
+
+pub fn is_valid_domain(domain: &str) -> bool {
+    if domain.is_empty() || domain.len() > 253 {
+        return false;
+    }
+    let labels: Vec<&str> = domain.split('.').collect();
+    if labels.len() < 2 {
+        return false;
+    }
+    for label in &labels {
+        if label.is_empty() || label.len() > 63 {
+            return false;
+        }
+        if !label.chars().all(|c| c.is_ascii_alphanumeric() || c == '-') {
+            return false;
+        }
+        if label.starts_with('-') || label.ends_with('-') {
+            return false;
+        }
+    }
+    true
+}