fix(replication): accept remote target healthCheckDuration nanoseconds (#5754)

* test(replication): accept madmin nanosecond healthCheckDuration payloads

Red-phase TDD tests for P0-7: mc 'replicate add' sends the madmin default
healthCheckDuration=60s as a Go time.Duration nanosecond integer
(60000000000), which RustFS currently rejects as an unsupported field and
would misread as seconds. Also pins the defensive seconds-or-nanos read
for persisted bucket-targets metadata and the capability contract listing
healthCheckDuration as writable.

Currently failing (red):
- remote_target_request_accepts_go_duration_wire_values
- remote_target_request_accepts_legacy_seconds_health_check
- remote_target_health_check_duration_is_declared_writable
- bucket_target_reads_go_nanosecond_durations_defensively
- runtime_capabilities_response_reports_missing_topology_before_storage_init

* fix(replication): accept remote target healthCheckDuration nanoseconds

mc 'replicate add' always sends the madmin default healthcheck-seconds=60
serialized as a Go time.Duration nanosecond integer (60000000000), so the
default mc link-creation path (and 'mc replicate update') failed with
InvalidRequest. Move healthCheckDuration from the unsupported to the
writable remote-target field list; the capability contract in the runtime
capabilities response follows the constants automatically.

Fix the unit mismatch in both directions:
- Request parsing and persisted bucket-targets reads decode the value
  defensively: below 10^7 it is legacy RustFS seconds, otherwise Go
  time.Duration nanoseconds (also covers MinIO-written metadata).
  totalDowntime shares the same wire shape and gets the same handling.
- The list-remote-targets admin response re-encodes only these two fields
  as nanoseconds via a dedicated serialization path, leaving the persisted
  seconds-based wire format untouched for existing readers.

The per-target health-check interval is accepted for mc compatibility but
not yet applied; the heartbeat keeps its global env-configured interval,
and the explicit 'healthcheck' update op stays rejected. disableProxy,
edge, and edgeSyncBeforeExpiry remain explicitly rejected.

crates/ecstore/src/api/mod.rs

@@ -206,7 +206,9 @@ pub mod bucket {
     }
 
     pub mod target {
-        pub use crate::bucket::target::{ARN, BucketTarget, BucketTargetType, BucketTargets, Credentials, LatencyStat};
+        pub use crate::bucket::target::{
+            ARN, BucketTarget, BucketTargetType, BucketTargets, Credentials, LatencyStat, duration_from_secs_or_nanos,
+        };
     }
 
     pub mod utils {

crates/ecstore/src/bucket/target/bucket_target.rs

@@ -94,6 +94,21 @@ mod duration_milliseconds {
     }
 }
 
+/// Defensive decode for the two integer wire encodings of these duration
+/// fields: RustFS persists (and legacy RustFS clients sent) plain seconds,
+/// while Go `time.Duration` JSON — madmin/mc requests and MinIO-written
+/// bucket-targets metadata — is nanoseconds. No meaningful interval lies
+/// between 10^7 seconds (~115 days) and 10^7 nanoseconds (10ms), so the
+/// magnitude disambiguates the unit.
+pub fn duration_from_secs_or_nanos(value: u64) -> Duration {
+    const NANOS_THRESHOLD: u64 = 10_000_000;
+    if value < NANOS_THRESHOLD {
+        Duration::from_secs(value)
+    } else {
+        Duration::from_nanos(value)
+    }
+}
+
 mod duration_seconds {
     use serde::{Deserialize, Deserializer, Serializer};
     use std::time::Duration;
@@ -109,8 +124,8 @@ mod duration_seconds {
     where
         D: Deserializer<'de>,
     {
-        let secs = u64::deserialize(deserializer)?;
-        Ok(Duration::from_secs(secs))
+        let value = u64::deserialize(deserializer)?;
+        Ok(super::duration_from_secs_or_nanos(value))
     }
 }
 
@@ -489,6 +504,29 @@ mod tests {
         assert_eq!(original.offline_count, deserialized.offline_count);
     }
 
+    #[test]
+    fn bucket_target_reads_go_nanosecond_durations_defensively() {
+        // MinIO-written bucket-targets metadata and madmin clients encode
+        // these fields as Go `time.Duration` nanoseconds; RustFS has always
+        // persisted seconds. Both encodings must decode to the same interval.
+        let target: BucketTarget = serde_json::from_value(serde_json::json!({
+            "endpoint": "localhost:9000",
+            "targetbucket": "target",
+            "type": "replication",
+            "healthCheckDuration": 60_000_000_000u64,
+            "totalDowntime": 90_000_000_000u64
+        }))
+        .expect("nanosecond durations should deserialize");
+
+        assert_eq!(target.health_check_duration, Duration::from_secs(60));
+        assert_eq!(target.total_downtime, Duration::from_secs(90));
+
+        // The persisted wire format stays seconds for existing RustFS readers.
+        let value = serde_json::to_value(&target).expect("target should serialize");
+        assert_eq!(value["healthCheckDuration"], 60);
+        assert_eq!(value["totalDowntime"], 90);
+    }
+
     #[test]
     fn test_bucket_target_debug_redacts_credentials() {
         let target = BucketTarget {

crates/replication/src/config.rs

@@ -73,9 +73,13 @@ pub const REMOTE_TARGET_WRITABLE_FIELDS: &[&str] = &[
     "storage_class",
     "skipTlsVerify",
     "caCertPem",
+    // Accepted for mc compatibility (mc `replicate add` always sends the
+    // madmin default of 60s); the per-target health-check interval is not
+    // yet applied — the heartbeat keeps its global env-configured interval.
+    "healthCheckDuration",
 ];
 
-pub const REMOTE_TARGET_UNSUPPORTED_FIELDS: &[&str] = &["disableProxy", "healthCheckDuration", "edge", "edgeSyncBeforeExpiry"];
+pub const REMOTE_TARGET_UNSUPPORTED_FIELDS: &[&str] = &["disableProxy", "edge", "edgeSyncBeforeExpiry"];
 
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
 pub struct ObjectOpts {

rustfs/src/admin/handlers/replication.rs

@@ -26,7 +26,9 @@ use crate::admin::storage_api::bucket::replication::REMOTE_TARGET_UNSUPPORTED_FI
 #[cfg(test)]
 use crate::admin::storage_api::bucket::replication::REMOTE_TARGET_WRITABLE_FIELDS;
 use crate::admin::storage_api::bucket::replication::{BucketStats, ReplicationStatusType};
-use crate::admin::storage_api::bucket::target::{BucketTarget, BucketTargetType, Credentials as TargetCredentials, LatencyStat};
+use crate::admin::storage_api::bucket::target::{
+    BucketTarget, BucketTargetType, Credentials as TargetCredentials, LatencyStat, duration_from_secs_or_nanos,
+};
 use crate::admin::storage_api::bucket::target_sys::{BucketTargetError, BucketTargetSys};
 use crate::admin::storage_api::contract::bucket::{BucketOperations, BucketOptions};
 use crate::admin::storage_api::contract::list::ListOperations as _;
@@ -295,12 +297,12 @@ impl RemoteTargetRequest {
             ));
         }
 
-        for (unsupported, configured) in REMOTE_TARGET_UNSUPPORTED_FIELDS.iter().copied().zip([
-            self.disable_proxy,
-            self.health_check_duration != 0,
-            self.edge,
-            self.edge_sync_before_expiry,
-        ]) {
+        for (unsupported, configured) in
+            REMOTE_TARGET_UNSUPPORTED_FIELDS
+                .iter()
+                .copied()
+                .zip([self.disable_proxy, self.edge, self.edge_sync_before_expiry])
+        {
             if configured {
                 return Err(s3_error!(
                     InvalidRequest,
@@ -325,11 +327,15 @@ impl RemoteTargetRequest {
             storage_class: self.storage_class,
             skip_tls_verify: self.skip_tls_verify,
             ca_cert_pem: self.ca_cert_pem,
-            health_check_duration: Duration::from_secs(self.health_check_duration),
+            // madmin/mc encode these Go `time.Duration` fields as nanoseconds;
+            // legacy RustFS clients sent seconds. Accepted for mc compatibility;
+            // the per-target health-check interval is not yet applied — the
+            // heartbeat keeps its global env-configured interval.
+            health_check_duration: duration_from_secs_or_nanos(self.health_check_duration),
             disable_proxy: self.disable_proxy,
             reset_before_date: self.reset_before_date,
             reset_id: self.reset_id,
-            total_downtime: Duration::from_secs(self.total_downtime),
+            total_downtime: duration_from_secs_or_nanos(self.total_downtime),
             last_online: self.last_online,
             online: self.online,
             latency: self.latency,
@@ -341,6 +347,23 @@ impl RemoteTargetRequest {
     }
 }
 
+/// Admin-response encoding of a remote target: the persisted bucket-targets
+/// format keeps `healthCheckDuration`/`totalDowntime` in seconds, but madmin
+/// decodes them as Go `time.Duration` (nanoseconds) — re-encode just those
+/// fields without touching the persistence wire format.
+fn remote_target_admin_json(target: &BucketTarget) -> Result<serde_json::Value, serde_json::Error> {
+    fn go_duration_nanos(duration: Duration) -> serde_json::Value {
+        // Saturate instead of truncating: >u64::MAX nanoseconds (~584 years)
+        // is unrepresentable for a Go time.Duration reader anyway.
+        u64::try_from(duration.as_nanos()).unwrap_or(u64::MAX).into()
+    }
+
+    let mut value = serde_json::to_value(target)?;
+    value["healthCheckDuration"] = go_duration_nanos(target.health_check_duration);
+    value["totalDowntime"] = go_duration_nanos(target.total_downtime);
+    Ok(value)
+}
+
 fn validate_remote_target_tls_settings(remote_target: &BucketTarget) -> S3Result<()> {
     let has_custom_ca = !remote_target.ca_cert_pem.trim().is_empty();
 
@@ -732,7 +755,14 @@ impl Operation for ListRemoteTargetHandler {
             let sys = BucketTargetSys::get();
             let targets = sys.list_targets(bucket, "").await;
 
-            let targets: Vec<_> = targets.iter().map(|target| target.redacted_credentials()).collect();
+            let targets: Vec<_> = targets
+                .iter()
+                .map(|target| remote_target_admin_json(&target.redacted_credentials()))
+                .collect::<Result<_, _>>()
+                .map_err(|e| {
+                    error!("Serialization error: {}", e);
+                    S3Error::with_message(S3ErrorCode::InternalError, "Failed to serialize targets".to_string())
+                })?;
             let json_targets = serde_json::to_vec(&targets).map_err(|e| {
                 error!("Serialization error: {}", e);
                 S3Error::with_message(S3ErrorCode::InternalError, "Failed to serialize targets".to_string())
@@ -1618,7 +1648,6 @@ mod tests {
             ("credentials.expiration", serde_json::json!("2026-01-01T00:00:00Z")),
             ("api", serde_json::json!("s3v2")),
             ("disableProxy", serde_json::json!(true)),
-            ("healthCheckDuration", serde_json::json!(5)),
             ("edge", serde_json::json!(true)),
             ("edgeSyncBeforeExpiry", serde_json::json!(true)),
         ] {
@@ -1679,6 +1708,65 @@ mod tests {
         assert!(!message.contains("secret"));
     }
 
+    #[test]
+    fn remote_target_request_accepts_go_duration_wire_values() {
+        // `mc replicate add` defaults `--healthcheck-seconds` to 60; madmin
+        // serializes Go `time.Duration` fields as nanosecond integers.
+        let mut request = valid_remote_target_request();
+        request["healthCheckDuration"] = serde_json::json!(60_000_000_000u64);
+        request["totalDowntime"] = serde_json::json!(90_000_000_000u64);
+
+        let target = serde_json::from_value::<RemoteTargetRequest>(request)
+            .expect("madmin-shaped request should deserialize")
+            .into_bucket_target()
+            .expect("mc default healthCheckDuration must be accepted");
+
+        assert_eq!(target.health_check_duration, std::time::Duration::from_secs(60));
+        assert_eq!(target.total_downtime, std::time::Duration::from_secs(90));
+    }
+
+    #[test]
+    fn remote_target_request_accepts_legacy_seconds_health_check() {
+        // Older RustFS clients sent these duration fields as plain seconds.
+        let mut request = valid_remote_target_request();
+        request["healthCheckDuration"] = serde_json::json!(60);
+
+        let target = serde_json::from_value::<RemoteTargetRequest>(request)
+            .expect("request should deserialize")
+            .into_bucket_target()
+            .expect("legacy seconds healthCheckDuration must be accepted");
+
+        assert_eq!(target.health_check_duration, std::time::Duration::from_secs(60));
+    }
+
+    #[test]
+    fn list_remote_targets_response_encodes_go_durations_as_nanoseconds() {
+        // madmin (and therefore mc) decode healthCheckDuration/totalDowntime
+        // as Go time.Duration nanoseconds; the persisted format stays seconds.
+        let target = BucketTarget {
+            endpoint: "192.168.1.10:9000".to_string(),
+            target_bucket: "target".to_string(),
+            health_check_duration: std::time::Duration::from_secs(60),
+            total_downtime: std::time::Duration::from_secs(90),
+            ..Default::default()
+        };
+
+        let value = super::remote_target_admin_json(&target).expect("admin response should serialize");
+
+        assert_eq!(value["healthCheckDuration"], 60_000_000_000u64);
+        assert_eq!(value["totalDowntime"], 90_000_000_000u64);
+        // Persistence keeps seconds: the response path must not leak into it.
+        let persisted = serde_json::to_value(&target).expect("persisted form should serialize");
+        assert_eq!(persisted["healthCheckDuration"], 60);
+        assert_eq!(persisted["totalDowntime"], 90);
+    }
+
+    #[test]
+    fn remote_target_health_check_duration_is_declared_writable() {
+        assert!(REMOTE_TARGET_WRITABLE_FIELDS.contains(&"healthCheckDuration"));
+        assert!(!REMOTE_TARGET_UNSUPPORTED_FIELDS.contains(&"healthCheckDuration"));
+    }
+
     #[test]
     fn remote_target_capability_fields_do_not_overlap() {
         for field in REMOTE_TARGET_UNSUPPORTED_FIELDS {

rustfs/src/admin/handlers/system.rs

@@ -1279,6 +1279,14 @@ mod tests {
                 .iter()
                 .any(|field| field.name == "disableProxy" && field.state == super::ReplicationFieldState::Unsupported)
         );
+        assert!(
+            response
+                .replication
+                .remote_targets
+                .fields
+                .iter()
+                .any(|field| field.name == "healthCheckDuration" && field.state == super::ReplicationFieldState::Supported)
+        );
         assert_eq!(response.manual_transition_jobs.contract_version, 1);
         assert_eq!(response.manual_transition_jobs.status.state, CapabilityState::Supported);
         assert_eq!(response.manual_transition_jobs.modes, ["enqueue_only", "async"]);
@@ -1361,6 +1369,13 @@ mod tests {
                 .iter()
                 .any(|field| field["name"] == "disableProxy" && field["state"] == "unsupported")
         );
+        assert!(
+            value["replication"]["remote_targets"]["fields"]
+                .as_array()
+                .expect("remote target fields should be an array")
+                .iter()
+                .any(|field| field["name"] == "healthCheckDuration" && field["state"] == "supported")
+        );
         assert_eq!(value["manual_transition_jobs"]["contract_version"], 1);
         assert_eq!(value["manual_transition_jobs"]["status"]["state"], "supported");
         assert_eq!(value["manual_transition_jobs"]["modes"], json!(["enqueue_only", "async"]));

rustfs/src/admin/storage_api.rs

@@ -599,6 +599,7 @@ pub(crate) mod replication {
 }
 
 pub(crate) mod target {
+    pub(crate) use super::ecstore_bucket::target::duration_from_secs_or_nanos;
     #[allow(clippy::upper_case_acronyms)]
     pub(crate) type ARN = super::ecstore_bucket::target::ARN;
     pub(crate) type BucketTarget = super::ecstore_bucket::target::BucketTarget;