feat(obs): improve metrics coverage and dashboard performance (#2682)

crates/common/src/internode_metrics.rs

@@ -51,7 +51,7 @@ impl InternodeMetrics {
             return;
         }
         self.sent_bytes_total.fetch_add(bytes, Ordering::Relaxed);
-        counter!("rustfs.internode.sent.bytes.total").increment(bytes);
+        counter!("rustfs_system_network_internode_sent_bytes_total").increment(bytes);
     }
 
     pub fn record_recv_bytes(&self, bytes: usize) {
@@ -60,22 +60,22 @@ impl InternodeMetrics {
             return;
         }
         self.recv_bytes_total.fetch_add(bytes, Ordering::Relaxed);
-        counter!("rustfs.internode.recv.bytes.total").increment(bytes);
+        counter!("rustfs_system_network_internode_recv_bytes_total").increment(bytes);
     }
 
     pub fn record_outgoing_request(&self) {
         self.outgoing_requests_total.fetch_add(1, Ordering::Relaxed);
-        counter!("rustfs.internode.requests.outgoing.total").increment(1);
+        counter!("rustfs_system_network_internode_requests_outgoing_total").increment(1);
     }
 
     pub fn record_incoming_request(&self) {
         self.incoming_requests_total.fetch_add(1, Ordering::Relaxed);
-        counter!("rustfs.internode.requests.incoming.total").increment(1);
+        counter!("rustfs_system_network_internode_requests_incoming_total").increment(1);
     }
 
     pub fn record_error(&self) {
         self.errors_total.fetch_add(1, Ordering::Relaxed);
-        counter!("rustfs.internode.errors.total").increment(1);
+        counter!("rustfs_system_network_internode_errors_total").increment(1);
     }
 
     pub fn record_dial_result(&self, duration: Duration, success: bool) {
@@ -83,11 +83,11 @@ impl InternodeMetrics {
         self.dial_total_time_nanos.fetch_add(elapsed_nanos, Ordering::Relaxed);
         let samples = self.dial_samples_total.fetch_add(1, Ordering::Relaxed) + 1;
         let total = self.dial_total_time_nanos.load(Ordering::Relaxed);
-        gauge!("rustfs.internode.dial.avg_time.nanos").set(total as f64 / samples as f64);
+        gauge!("rustfs_system_network_internode_dial_avg_time_nanos").set(total as f64 / samples as f64);
 
         if !success {
             self.dial_errors_total.fetch_add(1, Ordering::Relaxed);
-            counter!("rustfs.internode.dial.errors.total").increment(1);
+            counter!("rustfs_system_network_internode_dial_errors_total").increment(1);
         }
 
         let now_ms = SystemTime::now()

crates/ecstore/src/bucket/replication/replication_resyncer.rs

@@ -2578,6 +2578,7 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo {
             }
         };
 
+        let has_tagging_replication = !put_opts.user_tags.is_empty();
         if let Some(err) = if is_multipart {
             drop(gr);
             let result = replicate_object_with_multipart(MultipartReplicationContext {
@@ -2593,6 +2594,9 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo {
             })
             .await;
             record_proxy_request(&bucket, "PutObject", result.is_err()).await;
+            if has_tagging_replication {
+                record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await;
+            }
             result.err()
         } else {
             gr.stream = wrap_with_bandwidth_monitor(gr.stream, &put_opts, &bucket, &rinfo.arn);
@@ -2602,6 +2606,9 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo {
                 .await
                 .map_err(|e| std::io::Error::other(e.to_string()));
             record_proxy_request(&bucket, "PutObject", result.is_err()).await;
+            if has_tagging_replication {
+                record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await;
+            }
             result.err()
         } {
             rinfo.replication_status = ReplicationStatusType::Failed;
@@ -2867,6 +2874,7 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo {
                 }
             };
 
+            let has_tagging_replication = !put_opts.user_tags.is_empty();
             if let Some(err) = if is_multipart {
                 drop(gr);
                 let result = replicate_object_with_multipart(MultipartReplicationContext {
@@ -2882,6 +2890,9 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo {
                 })
                 .await;
                 record_proxy_request(&bucket, "PutObject", result.is_err()).await;
+                if has_tagging_replication {
+                    record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await;
+                }
                 result.err()
             } else {
                 gr.stream = wrap_with_bandwidth_monitor(gr.stream, &put_opts, &bucket, &rinfo.arn);
@@ -2891,6 +2902,9 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo {
                     .await
                     .map_err(|e| std::io::Error::other(e.to_string()));
                 record_proxy_request(&bucket, "PutObject", result.is_err()).await;
+                if has_tagging_replication {
+                    record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await;
+                }
                 result.err()
             } {
                 rinfo.replication_status = ReplicationStatusType::Failed;

crates/ecstore/src/bucket/replication/replication_state.rs

@@ -489,8 +489,14 @@ impl QueueCache {
 pub struct ProxyMetric {
     pub get_total: i64,
     pub get_failed: i64,
+    pub get_tag_total: i64,
+    pub get_tag_failed: i64,
     pub put_total: i64,
     pub put_failed: i64,
+    pub put_tag_total: i64,
+    pub put_tag_failed: i64,
+    pub delete_tag_total: i64,
+    pub delete_tag_failed: i64,
     pub head_total: i64,
     pub head_failed: i64,
 }
@@ -499,8 +505,14 @@ impl ProxyMetric {
     pub fn add(&mut self, other: &ProxyMetric) {
         self.get_total += other.get_total;
         self.get_failed += other.get_failed;
+        self.get_tag_total += other.get_tag_total;
+        self.get_tag_failed += other.get_tag_failed;
         self.put_total += other.put_total;
         self.put_failed += other.put_failed;
+        self.put_tag_total += other.put_tag_total;
+        self.put_tag_failed += other.put_tag_failed;
+        self.delete_tag_total += other.delete_tag_total;
+        self.delete_tag_failed += other.delete_tag_failed;
         self.head_total += other.head_total;
         self.head_failed += other.head_failed;
     }
@@ -527,18 +539,36 @@ impl ProxyStatsCache {
                     metric.get_failed += 1;
                 }
             }
+            "GetObjectTagging" => {
+                metric.get_tag_total += 1;
+                if is_err {
+                    metric.get_tag_failed += 1;
+                }
+            }
             "PutObject" => {
                 metric.put_total += 1;
                 if is_err {
                     metric.put_failed += 1;
                 }
             }
+            "PutObjectTagging" => {
+                metric.put_tag_total += 1;
+                if is_err {
+                    metric.put_tag_failed += 1;
+                }
+            }
             "HeadObject" => {
                 metric.head_total += 1;
                 if is_err {
                     metric.head_failed += 1;
                 }
             }
+            "DeleteObjectTagging" => {
+                metric.delete_tag_total += 1;
+                if is_err {
+                    metric.delete_tag_failed += 1;
+                }
+            }
             _ => {}
         }
     }

crates/ecstore/src/erasure_coding/encode.rs

@@ -26,6 +26,20 @@ use tokio::io::AsyncRead;
 use tokio::sync::mpsc;
 use tracing::error;
 
+const ENV_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BYTES: &str = "RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BYTES";
+const DEFAULT_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BYTES: usize = 32 * 1024 * 1024;
+const DEFAULT_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BLOCKS: usize = 8;
+
+fn encode_channel_capacity(expanded_block_bytes: usize, max_inflight_bytes: usize) -> usize {
+    if expanded_block_bytes == 0 {
+        return 1;
+    }
+
+    max_inflight_bytes
+        .saturating_div(expanded_block_bytes)
+        .clamp(1, DEFAULT_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BLOCKS)
+}
+
 pub(crate) struct MultiWriter<'a> {
     writers: &'a mut [Option<BitrotWriterWrapper>],
     write_quorum: usize,
@@ -185,7 +199,14 @@ impl Erasure {
     where
         R: AsyncRead + Send + Sync + Unpin + 'static,
     {
-        let (tx, mut rx) = mpsc::channel::<Vec<Bytes>>(8);
+        // Bound queued encoded blocks by memory budget to avoid per-request spikes.
+        let expanded_block_bytes = self.shard_size().saturating_mul(self.total_shard_count());
+        let max_inflight_bytes = rustfs_utils::get_env_usize(
+            ENV_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BYTES,
+            DEFAULT_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BYTES,
+        );
+        let inflight_blocks = encode_channel_capacity(expanded_block_bytes, max_inflight_bytes);
+        let (tx, mut rx) = mpsc::channel::<Vec<Bytes>>(inflight_blocks);
 
         let task = tokio::spawn(async move {
             let block_size = self.block_size;
@@ -310,4 +331,18 @@ mod tests {
         assert_eq!(written, b"small payload".len());
         assert!(!committed.lock().unwrap().is_empty());
     }
+
+    #[test]
+    fn encode_channel_capacity_never_returns_zero() {
+        assert_eq!(encode_channel_capacity(0, 1024), 1);
+        assert_eq!(encode_channel_capacity(4096, 0), 1);
+        assert_eq!(encode_channel_capacity(4096, 1024), 1);
+    }
+
+    #[test]
+    fn encode_channel_capacity_respects_budget_and_hard_cap() {
+        assert_eq!(encode_channel_capacity(4 * 1024 * 1024, 32 * 1024 * 1024), 8);
+        assert_eq!(encode_channel_capacity(16 * 1024 * 1024, 32 * 1024 * 1024), 2);
+        assert_eq!(encode_channel_capacity(1, usize::MAX), DEFAULT_RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BLOCKS);
+    }
 }

crates/iam/src/manager.rs

@@ -36,7 +36,7 @@ use rustfs_policy::{
 use rustfs_utils::{get_env_opt_str, path::path_join_buf};
 use serde::{Deserialize, Serialize};
 use serde_json::Value;
-use std::sync::atomic::AtomicU8;
+use std::sync::atomic::{AtomicU8, AtomicU64};
 use std::{
     collections::{HashMap, HashSet},
     sync::{
@@ -93,6 +93,17 @@ pub struct IamCache<T> {
     pub roles: HashMap<ARN, Vec<String>>,
     pub send_chan: Sender<i64>,
     pub last_timestamp: AtomicI64,
+    pub sync_failures: AtomicU64,
+    pub sync_successes: AtomicU64,
+    pub last_sync_duration_millis: AtomicU64,
+}
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct IamSyncMetricsSnapshot {
+    pub last_sync_duration_millis: u64,
+    pub since_last_sync_millis: u64,
+    pub sync_failures: u64,
+    pub sync_successes: u64,
 }
 
 impl<T> IamCache<T>
@@ -116,6 +127,9 @@ where
             send_chan: sender,
             roles: HashMap::new(),
             last_timestamp: AtomicI64::new(0),
+            sync_failures: AtomicU64::new(0),
+            sync_successes: AtomicU64::new(0),
+            last_sync_duration_millis: AtomicU64::new(0),
         });
 
         sys.clone().init(receiver).await.unwrap();
@@ -200,11 +214,38 @@ where
     }
 
     async fn load(self: Arc<Self>) -> Result<()> {
-        // debug!("load iam to cache");
-        self.api.load_all(&self.cache).await?;
-        self.last_timestamp
-            .store(OffsetDateTime::now_utc().unix_timestamp(), Ordering::Relaxed);
-        Ok(())
+        let started_at = std::time::Instant::now();
+        match self.api.load_all(&self.cache).await {
+            Ok(()) => {
+                self.last_timestamp
+                    .store(OffsetDateTime::now_utc().unix_timestamp(), Ordering::Relaxed);
+                self.sync_successes.fetch_add(1, Ordering::Relaxed);
+                self.last_sync_duration_millis
+                    .store(started_at.elapsed().as_millis().min(u128::from(u64::MAX)) as u64, Ordering::Relaxed);
+                Ok(())
+            }
+            Err(err) => {
+                self.sync_failures.fetch_add(1, Ordering::Relaxed);
+                Err(err)
+            }
+        }
+    }
+
+    pub fn sync_metrics_snapshot(&self) -> IamSyncMetricsSnapshot {
+        let now_secs = OffsetDateTime::now_utc().unix_timestamp();
+        let last_sync_secs = self.last_timestamp.load(Ordering::Relaxed);
+        let since_last_sync_millis = if last_sync_secs > 0 && now_secs >= last_sync_secs {
+            ((now_secs - last_sync_secs) as u64).saturating_mul(1000)
+        } else {
+            0
+        };
+
+        IamSyncMetricsSnapshot {
+            last_sync_duration_millis: self.last_sync_duration_millis.load(Ordering::Relaxed),
+            since_last_sync_millis,
+            sync_failures: self.sync_failures.load(Ordering::Relaxed),
+            sync_successes: self.sync_successes.load(Ordering::Relaxed),
+        }
     }
 
     pub async fn load_user(&self, access_key: &str) -> Result<()> {

crates/iam/src/oidc.rs

@@ -31,11 +31,11 @@ use rustfs_ecstore::config::{Config as ServerConfig, KVS, get_global_server_conf
 use rustfs_policy::policy::{ClaimLookup, get_claim_case_insensitive};
 use serde::{Deserialize, Serialize};
 use std::borrow::Cow;
-use std::collections::HashMap;
+use std::collections::{HashMap, VecDeque};
 use std::future::Future;
 use std::net::IpAddr;
 use std::pin::Pin;
-use std::sync::RwLock;
+use std::sync::{LazyLock, Mutex, MutexGuard, RwLock};
 use std::time::{Duration as StdDuration, Instant};
 use tokio::time::sleep;
 use tracing::{error, info, warn};
@@ -44,6 +44,127 @@ use url::Url;
 const OIDC_JWKS_REFRESH_INTERVAL: StdDuration = StdDuration::from_secs(24 * 60 * 60);
 const OIDC_DISCOVERY_TRANSPORT_RETRIES: usize = 3;
 const OIDC_DISCOVERY_TRANSPORT_RETRY_DELAY: StdDuration = StdDuration::from_millis(50);
+const OIDC_PLUGIN_AUTHN_WINDOW: StdDuration = StdDuration::from_secs(60);
+
+#[derive(Debug, Clone, Copy, Default)]
+pub struct OidcPluginAuthnMetricsSnapshot {
+    pub failed_requests_minute: u64,
+    pub last_fail_seconds: u64,
+    pub last_succ_seconds: u64,
+    pub succ_avg_rtt_ms_minute: u64,
+    pub succ_max_rtt_ms_minute: u64,
+    pub total_requests_minute: u64,
+}
+
+#[derive(Debug, Clone)]
+struct OidcPluginAuthnSample {
+    observed_at: Instant,
+    succeeded: bool,
+    rtt_ms: u64,
+}
+
+#[derive(Debug, Default)]
+struct OidcPluginAuthnMetrics {
+    samples: Mutex<VecDeque<OidcPluginAuthnSample>>,
+    last_fail_at: Mutex<Option<Instant>>,
+    last_succ_at: Mutex<Option<Instant>>,
+}
+
+fn lock_oidc_plugin_authn_metrics<'a, T>(mutex: &'a Mutex<T>, metric: &'static str) -> MutexGuard<'a, T> {
+    match mutex.lock() {
+        Ok(guard) => guard,
+        Err(err) => {
+            warn!("recovering poisoned OIDC plugin authn metrics lock: {}", metric);
+            err.into_inner()
+        }
+    }
+}
+
+fn seconds_since(now: Instant, observed_at: Option<Instant>) -> u64 {
+    observed_at
+        .map(|instant| now.duration_since(instant).as_secs())
+        .unwrap_or_default()
+}
+
+impl OidcPluginAuthnMetrics {
+    fn record(&self, rtt_ms: u64, succeeded: bool) {
+        let now = Instant::now();
+        let mut samples = lock_oidc_plugin_authn_metrics(&self.samples, "samples");
+        samples.push_back(OidcPluginAuthnSample {
+            observed_at: now,
+            succeeded,
+            rtt_ms,
+        });
+        while samples
+            .front()
+            .is_some_and(|sample| now.duration_since(sample.observed_at) > OIDC_PLUGIN_AUTHN_WINDOW)
+        {
+            samples.pop_front();
+        }
+        drop(samples);
+
+        if succeeded {
+            *lock_oidc_plugin_authn_metrics(&self.last_succ_at, "last_succ_at") = Some(now);
+        } else {
+            *lock_oidc_plugin_authn_metrics(&self.last_fail_at, "last_fail_at") = Some(now);
+        }
+    }
+
+    fn snapshot(&self) -> OidcPluginAuthnMetricsSnapshot {
+        let now = Instant::now();
+        let (total_requests_minute, failed_requests_minute, succ_avg_rtt_ms_minute, succ_max_rtt_ms_minute) = {
+            let mut samples = lock_oidc_plugin_authn_metrics(&self.samples, "samples");
+            while samples
+                .front()
+                .is_some_and(|sample| now.duration_since(sample.observed_at) > OIDC_PLUGIN_AUTHN_WINDOW)
+            {
+                samples.pop_front();
+            }
+
+            let mut failed_requests_minute = 0u64;
+            let mut successful_requests = 0u64;
+            let mut successful_rtt_sum = 0u64;
+            let mut succ_max_rtt_ms_minute = 0u64;
+
+            for sample in samples.iter() {
+                if sample.succeeded {
+                    successful_requests += 1;
+                    successful_rtt_sum += sample.rtt_ms;
+                    succ_max_rtt_ms_minute = succ_max_rtt_ms_minute.max(sample.rtt_ms);
+                } else {
+                    failed_requests_minute += 1;
+                }
+            }
+
+            let succ_avg_rtt_ms_minute = successful_rtt_sum.checked_div(successful_requests).unwrap_or_default();
+
+            (
+                samples.len() as u64,
+                failed_requests_minute,
+                succ_avg_rtt_ms_minute,
+                succ_max_rtt_ms_minute,
+            )
+        };
+
+        let last_fail_seconds = seconds_since(now, *lock_oidc_plugin_authn_metrics(&self.last_fail_at, "last_fail_at"));
+        let last_succ_seconds = seconds_since(now, *lock_oidc_plugin_authn_metrics(&self.last_succ_at, "last_succ_at"));
+
+        OidcPluginAuthnMetricsSnapshot {
+            failed_requests_minute,
+            last_fail_seconds,
+            last_succ_seconds,
+            succ_avg_rtt_ms_minute,
+            succ_max_rtt_ms_minute,
+            total_requests_minute,
+        }
+    }
+}
+
+static OIDC_PLUGIN_AUTHN_METRICS: LazyLock<OidcPluginAuthnMetrics> = LazyLock::new(OidcPluginAuthnMetrics::default);
+
+pub fn oidc_plugin_authn_metrics_snapshot() -> OidcPluginAuthnMetricsSnapshot {
+    OIDC_PLUGIN_AUTHN_METRICS.snapshot()
+}
 
 // ---- HTTP Client Adapter ----
 
@@ -120,6 +241,7 @@ impl<'c> AsyncHttpClient<'c> for ReqwestHttpClient {
 
     fn call(&'c self, request: http::Request<Vec<u8>>) -> Self::Future {
         Box::pin(async move {
+            let started_at = Instant::now();
             let (parts, body) = request.into_parts();
             let uri = parts.uri.to_string();
             let client = self.client_for_uri(&uri);
@@ -128,8 +250,13 @@ impl<'c> AsyncHttpClient<'c> for ReqwestHttpClient {
                 .headers(parts.headers)
                 .body(body)
                 .send()
-                .await
-                .map_err(OidcHttpError::Reqwest)?;
+                .await;
+
+            let elapsed_ms = started_at.elapsed().as_millis().min(u128::from(u64::MAX)) as u64;
+            let succeeded = response.as_ref().is_ok_and(|resp| resp.status().is_success());
+            OIDC_PLUGIN_AUTHN_METRICS.record(elapsed_ms, succeeded);
+
+            let response = response.map_err(OidcHttpError::Reqwest)?;
 
             let status = response.status();
             let headers = response.headers().clone();

crates/iam/src/sys.rs

@@ -16,9 +16,9 @@ use crate::error::Error as IamError;
 use crate::error::is_err_no_such_account;
 use crate::error::is_err_no_such_temp_account;
 use crate::error::{Error, Result};
-use crate::manager::IamCache;
 use crate::manager::extract_jwt_claims;
 use crate::manager::get_default_policyes;
+use crate::manager::{IamCache, IamSyncMetricsSnapshot};
 use crate::store::GroupInfo;
 use crate::store::MappedPolicy;
 use crate::store::Store;
@@ -186,6 +186,10 @@ impl<T: Store> IamSys<T> {
         self.store.api.has_watcher()
     }
 
+    pub fn sync_metrics_snapshot(&self) -> IamSyncMetricsSnapshot {
+        self.store.sync_metrics_snapshot()
+    }
+
     pub async fn set_policy_plugin_client(client: rustfs_policy::policy::opa::AuthZPlugin) {
         let policy_plugin_client = get_policy_plugin_client();
         let mut guard = policy_plugin_client.write().await;

crates/io-metrics/src/adaptive_ttl.rs

@@ -31,14 +31,14 @@ use std::time::{Duration, Instant};
 pub fn record_ttl_adjustment(_key: &str, base_ttl: u64, adjusted_ttl: u64) {
     use metrics::{counter, gauge};
 
-    counter!("rustfs.cache.ttl.adjustments").increment(1);
-    gauge!("rustfs.cache.ttl.base").set(base_ttl as f64);
-    gauge!("rustfs.cache.ttl.adjusted").set(adjusted_ttl as f64);
+    counter!("rustfs_cache_ttl_adjustments").increment(1);
+    gauge!("rustfs_cache_ttl_base").set(base_ttl as f64);
+    gauge!("rustfs_cache_ttl_adjusted").set(adjusted_ttl as f64);
 
     if adjusted_ttl > base_ttl {
-        counter!("rustfs.cache.ttl.extensions").increment(1);
+        counter!("rustfs_cache_ttl_extensions").increment(1);
     } else if adjusted_ttl < base_ttl {
-        counter!("rustfs.cache.ttl.reductions").increment(1);
+        counter!("rustfs_cache_ttl_reductions").increment(1);
     }
 }
 
@@ -46,7 +46,7 @@ pub fn record_ttl_adjustment(_key: &str, base_ttl: u64, adjusted_ttl: u64) {
 #[inline(always)]
 pub fn record_ttl_expiration() {
     use metrics::counter;
-    counter!("rustfs.cache.ttl.expirations").increment(1);
+    counter!("rustfs_cache_ttl_expirations").increment(1);
 }
 
 /// Record early eviction.
@@ -57,7 +57,7 @@ pub fn record_ttl_expiration() {
 #[inline(always)]
 pub fn record_early_eviction(reason: &str) {
     use metrics::counter;
-    counter!("rustfs.cache.evictions.early", "reason" => reason.to_string()).increment(1);
+    counter!("rustfs_cache_evictions_early", "reason" => reason.to_string()).increment(1);
 }
 
 /// Record access pattern change.
@@ -69,7 +69,7 @@ pub fn record_early_eviction(reason: &str) {
 #[inline(always)]
 pub fn record_access_pattern_change(from: &str, to: &str) {
     use metrics::counter;
-    counter!("rustfs.cache.access_pattern.changes", "from" => from.to_string(), "to" => to.to_string()).increment(1);
+    counter!("rustfs_cache_access_pattern_changes", "from" => from.to_string(), "to" => to.to_string()).increment(1);
 }
 
 /// Adaptive TTL statistics.

crates/io-metrics/src/backpressure_metrics.rs

@@ -18,35 +18,35 @@
 #[inline(always)]
 pub fn record_backpressure_state_change(from: &str, to: &str) {
     use metrics::counter;
-    counter!("rustfs.backpressure.state.changes", "from" => from.to_string(), "to" => to.to_string()).increment(1);
+    counter!("rustfs_backpressure_state_changes", "from" => from.to_string(), "to" => to.to_string()).increment(1);
 }
 
 /// Record backpressure rejection.
 #[inline(always)]
 pub fn record_backpressure_rejection() {
     use metrics::counter;
-    counter!("rustfs.backpressure.rejections").increment(1);
+    counter!("rustfs_backpressure_rejections").increment(1);
 }
 
 /// Record concurrent operations count.
 #[inline(always)]
 pub fn record_concurrent_operations(count: usize) {
     use metrics::gauge;
-    gauge!("rustfs.backpressure.concurrent").set(count as f64);
+    gauge!("rustfs_backpressure_concurrent").set(count as f64);
 }
 
 /// Record backpressure activation.
 #[inline(always)]
 pub fn record_backpressure_activation() {
     use metrics::counter;
-    counter!("rustfs.backpressure.activations").increment(1);
+    counter!("rustfs_backpressure_activations").increment(1);
 }
 
 /// Record backpressure deactivation.
 #[inline(always)]
 pub fn record_backpressure_deactivation() {
     use metrics::counter;
-    counter!("rustfs.backpressure.deactivations").increment(1);
+    counter!("rustfs_backpressure_deactivations").increment(1);
 }
 
 #[cfg(test)]

crates/io-metrics/src/capacity_metrics.rs

@@ -20,35 +20,35 @@ use std::time::Duration;
 /// Record capacity cache hit.
 #[inline(always)]
 pub fn record_capacity_cache_hit() {
-    counter!("rustfs.capacity.cache.hits").increment(1);
+    counter!("rustfs_capacity_cache_hits").increment(1);
 }
 
 /// Record capacity cache miss.
 #[inline(always)]
 pub fn record_capacity_cache_miss() {
-    counter!("rustfs.capacity.cache.misses").increment(1);
+    counter!("rustfs_capacity_cache_misses").increment(1);
 }
 
 /// Record how capacity cache was served to the caller.
 #[inline(always)]
 pub fn record_capacity_cache_served(state: &'static str) {
-    counter!("rustfs.capacity.cache.served.total", "state" => state).increment(1);
+    counter!("rustfs_capacity_cache_served_total", "state" => state).increment(1);
 }
 
 /// Record current capacity gauge.
 #[inline(always)]
 pub fn record_capacity_current_bytes(used_bytes: u64) {
-    gauge!("rustfs.capacity.current").set(used_bytes as f64);
+    gauge!("rustfs_capacity_current_bytes").set(used_bytes as f64);
 }
 
 /// Record capacity update completion.
 #[inline(always)]
 pub fn record_capacity_update_completed(source: &'static str, duration: Duration, used_bytes: u64, is_estimated: bool) {
-    counter!("rustfs.capacity.update.total", "source" => source).increment(1);
-    histogram!("rustfs.capacity.update.duration.seconds", "source" => source).record(duration.as_secs_f64());
-    histogram!("rustfs.capacity.update.bytes", "source" => source).record(used_bytes as f64);
+    counter!("rustfs_capacity_update_total", "source" => source).increment(1);
+    histogram!("rustfs_capacity_update_duration_seconds", "source" => source).record(duration.as_secs_f64());
+    histogram!("rustfs_capacity_update_bytes", "source" => source).record(used_bytes as f64);
     counter!(
-        "rustfs.capacity.update.estimated.total",
+        "rustfs_capacity_update_estimated_total",
         "source" => source,
         "estimated" => if is_estimated { "true" } else { "false" }
     )
@@ -58,86 +58,86 @@ pub fn record_capacity_update_completed(source: &'static str, duration: Duration
 /// Record failed capacity update.
 #[inline(always)]
 pub fn record_capacity_update_failed(source: &'static str) {
-    counter!("rustfs.capacity.update.failures", "source" => source).increment(1);
+    counter!("rustfs_capacity_update_failures", "source" => source).increment(1);
 }
 
 /// Record a capacity refresh request.
 #[inline(always)]
 pub fn record_capacity_refresh_request(mode: &'static str, source: &'static str) {
-    counter!("rustfs.capacity.refresh.requests.total", "mode" => mode, "source" => source).increment(1);
+    counter!("rustfs_capacity_refresh_requests_total", "mode" => mode, "source" => source).increment(1);
 }
 
 /// Record a refresh joiner waiting for an inflight refresh.
 #[inline(always)]
 pub fn record_capacity_refresh_joiner(source: &'static str) {
-    counter!("rustfs.capacity.refresh.joiners.total", "source" => source).increment(1);
+    counter!("rustfs_capacity_refresh_joiners_total", "source" => source).increment(1);
 }
 
 /// Record the number of inflight capacity refreshes.
 #[inline(always)]
 pub fn record_capacity_refresh_inflight(count: usize) {
-    gauge!("rustfs.capacity.refresh.inflight").set(count as f64);
+    gauge!("rustfs_capacity_refresh_inflight").set(count as f64);
 }
 
 /// Record the final result of a capacity refresh.
 #[inline(always)]
 pub fn record_capacity_refresh_result(source: &'static str, result: &'static str, duration: Duration) {
-    counter!("rustfs.capacity.refresh.result.total", "source" => source, "result" => result).increment(1);
-    histogram!("rustfs.capacity.refresh.duration.seconds", "source" => source, "result" => result).record(duration.as_secs_f64());
+    counter!("rustfs_capacity_refresh_result_total", "source" => source, "result" => result).increment(1);
+    histogram!("rustfs_capacity_refresh_duration_seconds", "source" => source, "result" => result).record(duration.as_secs_f64());
 }
 
 /// Record the refresh scope selected for a capacity refresh.
 #[inline(always)]
 pub fn record_capacity_refresh_scope(scope: &'static str, disk_count: usize) {
-    counter!("rustfs.capacity.refresh.scope.total", "scope" => scope).increment(1);
-    histogram!("rustfs.capacity.refresh.scope.disks", "scope" => scope).record(disk_count as f64);
+    counter!("rustfs_capacity_refresh_scope_total", "scope" => scope).increment(1);
+    histogram!("rustfs_capacity_refresh_scope_disks", "scope" => scope).record(disk_count as f64);
 }
 
 /// Record the current number of dirty disks tracked by capacity management.
 #[inline(always)]
 pub fn record_capacity_dirty_disk_count(count: usize) {
-    gauge!("rustfs.capacity.dirty.disks").set(count as f64);
+    gauge!("rustfs_capacity_dirty_disks").set(count as f64);
 }
 
 /// Record capacity write activity.
 #[inline(always)]
 pub fn record_capacity_write_operation(write_frequency: usize) {
-    counter!("rustfs.capacity.write.operations").increment(1);
-    gauge!("rustfs.capacity.write.frequency").set(write_frequency as f64);
+    counter!("rustfs_capacity_write_operations").increment(1);
+    gauge!("rustfs_capacity_write_frequency").set(write_frequency as f64);
 }
 
 /// Record symlink accounting.
 #[inline(always)]
 pub fn record_capacity_symlink(size_bytes: u64) {
-    counter!("rustfs.capacity.symlinks.encountered").increment(1);
-    histogram!("rustfs.capacity.symlinks.size.bytes").record(size_bytes as f64);
+    counter!("rustfs_capacity_symlinks_encountered").increment(1);
+    histogram!("rustfs_capacity_symlinks_size_bytes").record(size_bytes as f64);
 }
 
 /// Record timeout fallback event.
 #[inline(always)]
 pub fn record_capacity_timeout_fallback() {
-    counter!("rustfs.capacity.timeout.fallback").increment(1);
+    counter!("rustfs_capacity_timeout_fallback").increment(1);
 }
 
 /// Record stall detection event.
 #[inline(always)]
 pub fn record_capacity_stall_detected() {
-    counter!("rustfs.capacity.timeout.stall").increment(1);
+    counter!("rustfs_capacity_timeout_stall").increment(1);
 }
 
 /// Record dynamic timeout usage.
 #[inline(always)]
 pub fn record_capacity_dynamic_timeout(timeout: Duration) {
-    counter!("rustfs.capacity.timeout.dynamic").increment(1);
-    histogram!("rustfs.capacity.timeout.dynamic.seconds").record(timeout.as_secs_f64());
+    counter!("rustfs_capacity_timeout_dynamic").increment(1);
+    histogram!("rustfs_capacity_timeout_dynamic_seconds").record(timeout.as_secs_f64());
 }
 
 /// Record scan sampling outcome.
 #[inline(always)]
 pub fn record_capacity_scan_sampling(sampled_count: usize, estimated: bool) {
-    histogram!("rustfs.capacity.scan.sampled.count").record(sampled_count as f64);
+    histogram!("rustfs_capacity_scan_sampled_count").record(sampled_count as f64);
     counter!(
-        "rustfs.capacity.scan.estimated.total",
+        "rustfs_capacity_scan_estimated_total",
         "estimated" => if estimated { "true" } else { "false" }
     )
     .increment(1);
@@ -146,7 +146,7 @@ pub fn record_capacity_scan_sampling(sampled_count: usize, estimated: bool) {
 /// Record the scan mode used for a capacity result.
 #[inline(always)]
 pub fn record_capacity_scan_mode(mode: &'static str) {
-    counter!("rustfs.capacity.scan.mode.total", "mode" => mode).increment(1);
+    counter!("rustfs_capacity_scan_mode_total", "mode" => mode).increment(1);
 }
 
 /// Record per-disk capacity scan statistics.
@@ -159,16 +159,16 @@ pub fn record_capacity_scan_disk(
     estimated: bool,
     partial_errors: bool,
 ) {
-    histogram!("rustfs.capacity.scan.disk.duration.seconds", "disk" => disk.to_owned()).record(duration.as_secs_f64());
-    histogram!("rustfs.capacity.scan.disk.files", "disk" => disk.to_owned()).record(file_count as f64);
-    histogram!("rustfs.capacity.scan.disk.sampled", "disk" => disk.to_owned()).record(sampled_count as f64);
+    histogram!("rustfs_capacity_scan_disk_duration_seconds", "disk" => disk.to_owned()).record(duration.as_secs_f64());
+    histogram!("rustfs_capacity_scan_disk_files", "disk" => disk.to_owned()).record(file_count as f64);
+    histogram!("rustfs_capacity_scan_disk_sampled", "disk" => disk.to_owned()).record(sampled_count as f64);
     counter!(
-        "rustfs.capacity.scan.disk.estimated.total",
+        "rustfs_capacity_scan_disk_estimated_total",
         "disk" => disk.to_owned(),
         "estimated" => if estimated { "true" } else { "false" }
     )
     .increment(1);
     if partial_errors {
-        counter!("rustfs.capacity.scan.disk.partial_errors.total", "disk" => disk.to_owned()).increment(1);
+        counter!("rustfs_capacity_scan_disk_partial_errors_total", "disk" => disk.to_owned()).increment(1);
     }
 }

crates/io-metrics/src/deadlock_metrics.rs

@@ -20,52 +20,52 @@ use std::time::Duration;
 #[inline(always)]
 pub fn record_deadlock_detected(cycle_length: usize) {
     use metrics::{counter, histogram};
-    counter!("rustfs.deadlock.detected").increment(1);
-    histogram!("rustfs.deadlock.cycle_length").record(cycle_length as f64);
+    counter!("rustfs_deadlock_detected_total").increment(1);
+    histogram!("rustfs_deadlock_cycle_length").record(cycle_length as f64);
 }
 
 /// Record long-held lock.
 #[inline(always)]
 pub fn record_long_held_lock(_lock_id: u64, hold_time: Duration) {
     use metrics::{counter, histogram};
-    counter!("rustfs.deadlock.long_held").increment(1);
-    histogram!("rustfs.deadlock.hold_time.secs").record(hold_time.as_secs_f64());
+    counter!("rustfs_deadlock_long_held").increment(1);
+    histogram!("rustfs_deadlock_hold_time_secs").record(hold_time.as_secs_f64());
 }
 
 /// Record lock acquisition.
 #[inline(always)]
 pub fn record_lock_acquisition(lock_type: &str) {
     use metrics::counter;
-    counter!("rustfs.lock.acquisitions", "type" => lock_type.to_string()).increment(1);
+    counter!("rustfs_lock_acquisitions", "type" => lock_type.to_string()).increment(1);
 }
 
 /// Record lock release.
 #[inline(always)]
 pub fn record_lock_release(lock_type: &str, hold_time: Duration) {
     use metrics::{counter, histogram};
-    counter!("rustfs.lock.releases", "type" => lock_type.to_string()).increment(1);
-    histogram!("rustfs.lock.hold_time.secs", "type" => lock_type.to_string()).record(hold_time.as_secs_f64());
+    counter!("rustfs_lock_releases", "type" => lock_type.to_string()).increment(1);
+    histogram!("rustfs_lock_hold_time_secs", "type" => lock_type.to_string()).record(hold_time.as_secs_f64());
 }
 
 /// Record lock contention.
 #[inline(always)]
 pub fn record_lock_contention(lock_type: &str) {
     use metrics::counter;
-    counter!("rustfs.lock.contentions", "type" => lock_type.to_string()).increment(1);
+    counter!("rustfs_lock_contentions", "type" => lock_type.to_string()).increment(1);
 }
 
 /// Record wait graph edge added.
 #[inline(always)]
 pub fn record_wait_edge_added() {
     use metrics::counter;
-    counter!("rustfs.deadlock.wait_edges.added").increment(1);
+    counter!("rustfs_deadlock_wait_edges_added").increment(1);
 }
 
 /// Record wait graph edge removed.
 #[inline(always)]
 pub fn record_wait_edge_removed() {
     use metrics::counter;
-    counter!("rustfs.deadlock.wait_edges.removed").increment(1);
+    counter!("rustfs_deadlock_wait_edges_removed").increment(1);
 }
 
 #[cfg(test)]

crates/io-metrics/src/io_metrics.rs

@@ -27,11 +27,11 @@
 pub fn record_io_scheduler_decision(buffer_size: usize, load_level: &str, strategy: &str) {
     use metrics::{counter, gauge, histogram};
 
-    counter!("rustfs.io.scheduler.decisions").increment(1);
-    gauge!("rustfs.io.scheduler.buffer_size").set(buffer_size as f64);
-    counter!("rustfs.io.scheduler.load", "level" => load_level.to_string()).increment(1);
-    counter!("rustfs.io.scheduler.strategy", "type" => strategy.to_string()).increment(1);
-    histogram!("rustfs.io.scheduler.buffer_size.histogram").record(buffer_size as f64);
+    counter!("rustfs_io_scheduler_decisions").increment(1);
+    gauge!("rustfs_io_scheduler_buffer_size").set(buffer_size as f64);
+    counter!("rustfs_io_scheduler_load", "level" => load_level.to_string()).increment(1);
+    counter!("rustfs_io_scheduler_strategy", "type" => strategy.to_string()).increment(1);
+    histogram!("rustfs_io_scheduler_buffer_size_histogram").record(buffer_size as f64);
 }
 
 /// Record I/O priority decision.
@@ -44,9 +44,9 @@ pub fn record_io_scheduler_decision(buffer_size: usize, load_level: &str, strate
 pub fn record_io_priority_decision(priority: &str, size: usize) {
     use metrics::{counter, histogram};
 
-    counter!("rustfs.io.priority.decisions").increment(1);
-    counter!("rustfs.io.priority.by_level", "priority" => priority.to_string()).increment(1);
-    histogram!("rustfs.io.priority.request_size").record(size as f64);
+    counter!("rustfs_io_priority_decisions").increment(1);
+    counter!("rustfs_io_priority_by_level", "priority" => priority.to_string()).increment(1);
+    histogram!("rustfs_io_priority_request_size").record(size as f64);
 }
 
 /// Record load level change.
@@ -58,7 +58,7 @@ pub fn record_io_priority_decision(priority: &str, size: usize) {
 #[inline(always)]
 pub fn record_load_level_change(from: &str, to: &str) {
     use metrics::counter;
-    counter!("rustfs.io.load.changes", "from" => from.to_string(), "to" => to.to_string()).increment(1);
+    counter!("rustfs_io_load_changes", "from" => from.to_string(), "to" => to.to_string()).increment(1);
 }
 
 /// Record bandwidth observation.
@@ -69,8 +69,8 @@ pub fn record_load_level_change(from: &str, to: &str) {
 #[inline(always)]
 pub fn record_bandwidth_observation(bps: u64) {
     use metrics::{gauge, histogram};
-    gauge!("rustfs.io.bandwidth.bps").set(bps as f64);
-    histogram!("rustfs.io.bandwidth.histogram").record(bps as f64);
+    gauge!("rustfs_io_bandwidth_bps").set(bps as f64);
+    histogram!("rustfs_io_bandwidth_histogram").record(bps as f64);
 }
 
 /// Record buffer size adjustment.
@@ -83,9 +83,9 @@ pub fn record_bandwidth_observation(bps: u64) {
 #[inline(always)]
 pub fn record_buffer_size_adjustment(original: usize, adjusted: usize, reason: &str) {
     use metrics::{counter, gauge};
-    counter!("rustfs.io.buffer.adjustments", "reason" => reason.to_string()).increment(1);
-    gauge!("rustfs.io.buffer.original").set(original as f64);
-    gauge!("rustfs.io.buffer.adjusted").set(adjusted as f64);
+    counter!("rustfs_io_buffer_adjustments", "reason" => reason.to_string()).increment(1);
+    gauge!("rustfs_io_buffer_original").set(original as f64);
+    gauge!("rustfs_io_buffer_adjusted").set(adjusted as f64);
 }
 
 /// Record queue operation.
@@ -98,8 +98,8 @@ pub fn record_buffer_size_adjustment(original: usize, adjusted: usize, reason: &
 #[inline(always)]
 pub fn record_queue_operation(operation: &str, priority: &str, queue_size: usize) {
     use metrics::{counter, gauge};
-    counter!("rustfs.io.queue.operations", "operation" => operation.to_string(), "priority" => priority.to_string()).increment(1);
-    gauge!("rustfs.io.queue.size", "priority" => priority.to_string()).set(queue_size as f64);
+    counter!("rustfs_io_queue_operations", "operation" => operation.to_string(), "priority" => priority.to_string()).increment(1);
+    gauge!("rustfs_io_queue_size", "priority" => priority.to_string()).set(queue_size as f64);
 }
 
 /// Record starvation event.
@@ -110,7 +110,7 @@ pub fn record_queue_operation(operation: &str, priority: &str, queue_size: usize
 #[inline(always)]
 pub fn record_starvation_event(priority: &str) {
     use metrics::counter;
-    counter!("rustfs.io.starvation.events", "priority" => priority.to_string()).increment(1);
+    counter!("rustfs_io_starvation_events", "priority" => priority.to_string()).increment(1);
 }
 
 /// I/O scheduler statistics.

crates/io-metrics/src/lib.rs

@@ -217,9 +217,9 @@ pub fn record_get_object_io_state(
 /// * `duration_ms` - Time taken for the read operation in milliseconds
 #[inline(always)]
 pub fn record_zero_copy_read(size_bytes: usize, duration_ms: f64) {
-    counter!("rustfs.zero_copy.reads.total").increment(1);
-    histogram!("rustfs.zero_copy.read.size.bytes").record(size_bytes as f64);
-    histogram!("rustfs.zero_copy.read.duration.ms").record(duration_ms);
+    counter!("rustfs_zero_copy_reads_total").increment(1);
+    histogram!("rustfs_zero_copy_read_size_bytes").record(size_bytes as f64);
+    histogram!("rustfs_zero_copy_read_duration_ms").record(duration_ms);
 }
 
 /// Record memory copies avoided by using zero-copy.
@@ -229,7 +229,7 @@ pub fn record_zero_copy_read(size_bytes: usize, duration_ms: f64) {
 /// * `bytes_saved` - Number of bytes that would have been copied without zero-copy
 #[inline(always)]
 pub fn record_memory_copy_saved(bytes_saved: usize) {
-    counter!("rustfs.zero_copy.memory.saved.bytes").increment(bytes_saved as u64);
+    counter!("rustfs_zero_copy_memory_saved_bytes_total").increment(bytes_saved as u64);
 }
 
 /// Record a fallback from zero-copy to regular read.
@@ -242,7 +242,7 @@ pub fn record_memory_copy_saved(bytes_saved: usize) {
 /// * `reason` - Reason for the fallback (e.g., "mmap_unavailable", "file_too_large")
 #[inline(always)]
 pub fn record_zero_copy_fallback(reason: &str) {
-    counter!("rustfs.zero_copy.fallback.total", "reason" => reason.to_string()).increment(1);
+    counter!("rustfs_zero_copy_fallback_total", "reason" => reason.to_string()).increment(1);
 }
 
 // ============================================================================
@@ -258,13 +258,13 @@ pub fn record_zero_copy_fallback(reason: &str) {
 /// * `from_pool` - Whether buffer was reused from pool
 #[inline(always)]
 pub fn record_bytes_pool_acquire(tier: &str, size: usize, from_pool: bool) {
-    counter!("rustfs.bytes.pool.acquisitions.total", "tier" => tier.to_string()).increment(1);
-    gauge!("rustfs.bytes.pool.size.bytes", "tier" => tier.to_string()).set(size as f64);
+    counter!("rustfs_bytes_pool_acquisitions_total", "tier" => tier.to_string()).increment(1);
+    gauge!("rustfs_bytes_pool_size_bytes", "tier" => tier.to_string()).set(size as f64);
 
     if from_pool {
-        counter!("rustfs.bytes.pool.hits.total", "tier" => tier.to_string()).increment(1);
+        counter!("rustfs_bytes_pool_hits_total", "tier" => tier.to_string()).increment(1);
     } else {
-        counter!("rustfs.bytes.pool.misses.total", "tier" => tier.to_string()).increment(1);
+        counter!("rustfs_bytes_pool_misses_total", "tier" => tier.to_string()).increment(1);
     }
 }
 
@@ -275,7 +275,7 @@ pub fn record_bytes_pool_acquire(tier: &str, size: usize, from_pool: bool) {
 /// * `tier` - Pool tier ("small", "medium", "large", "xlarge")
 #[inline(always)]
 pub fn record_bytes_pool_return(tier: &str) {
-    counter!("rustfs.bytes.pool.returns.total", "tier" => tier.to_string()).increment(1);
+    counter!("rustfs_bytes_pool_returns_total", "tier" => tier.to_string()).increment(1);
 }
 
 /// Record current BytesPool allocated bytes.
@@ -286,7 +286,7 @@ pub fn record_bytes_pool_return(tier: &str) {
 /// * `bytes` - Currently allocated bytes
 #[inline(always)]
 pub fn record_bytes_pool_allocated(tier: &str, bytes: u64) {
-    gauge!("rustfs.bytes.pool.allocated.bytes", "tier" => tier.to_string()).set(bytes as f64);
+    gauge!("rustfs_bytes_pool_allocated_bytes", "tier" => tier.to_string()).set(bytes as f64);
 }
 
 /// Get BytesPool hit rate as a gauge metric.
@@ -297,7 +297,7 @@ pub fn record_bytes_pool_allocated(tier: &str, bytes: u64) {
 /// * `hit_rate` - Hit rate (0.0 - 1.0)
 #[inline(always)]
 pub fn record_bytes_pool_hit_rate(tier: &str, hit_rate: f64) {
-    gauge!("rustfs.bytes.pool.hit.rate", "tier" => tier.to_string()).set(hit_rate * 100.0);
+    gauge!("rustfs_bytes_pool_hit_rate", "tier" => tier.to_string()).set(hit_rate * 100.0);
 }
 
 /// Record zero-copy write operation.
@@ -308,9 +308,9 @@ pub fn record_bytes_pool_hit_rate(tier: &str, hit_rate: f64) {
 /// * `duration_ms` - Time taken for the write operation in milliseconds
 #[inline(always)]
 pub fn record_zero_copy_write(size_bytes: usize, duration_ms: f64) {
-    counter!("rustfs.zero_copy.write.total").increment(1);
-    histogram!("rustfs.zero_copy.write.size.bytes").record(size_bytes as f64);
-    histogram!("rustfs.zero_copy.write.duration.ms").record(duration_ms);
+    counter!("rustfs_zero_copy_write_total").increment(1);
+    histogram!("rustfs_zero_copy_write_size_bytes").record(size_bytes as f64);
+    histogram!("rustfs_zero_copy_write_duration_ms").record(duration_ms);
 }
 
 /// Record zero-copy write fallback.
@@ -322,7 +322,7 @@ pub fn record_zero_copy_write(size_bytes: usize, duration_ms: f64) {
 /// * `reason` - Reason for the fallback
 #[inline(always)]
 pub fn record_zero_copy_write_fallback(reason: &str) {
-    counter!("rustfs.zero_copy.write.fallback.total", "reason" => reason.to_string()).increment(1);
+    counter!("rustfs_zero_copy_write_fallback_total", "reason" => reason.to_string()).increment(1);
 }
 
 /// Record bytes saved from zero-copy.
@@ -332,7 +332,7 @@ pub fn record_zero_copy_write_fallback(reason: &str) {
 /// * `size_bytes` - Number of bytes saved from zero-copy
 #[inline(always)]
 pub fn record_bytes_saved(size_bytes: usize) {
-    counter!("rustfs.zero_copy.bytes.saved.total").increment(size_bytes as u64);
+    counter!("rustfs_zero_copy_bytes_saved_total").increment(size_bytes as u64);
 }
 
 // ============================================================================
@@ -350,11 +350,11 @@ pub fn record_bytes_saved(size_bytes: usize) {
 /// interpreted as the definitive source of truth for data-plane copy mode.
 #[inline(always)]
 pub fn record_get_object(duration_ms: f64, size_bytes: i64) {
-    counter!("rustfs.s3.get_object.total").increment(1);
-    histogram!("rustfs.s3.get_object.duration.ms").record(duration_ms);
+    counter!("rustfs_s3_get_object_total").increment(1);
+    histogram!("rustfs_s3_get_object_duration_ms").record(duration_ms);
 
     if size_bytes > 0 {
-        histogram!("rustfs.s3.get_object.size.bytes").record(size_bytes as f64);
+        histogram!("rustfs_s3_get_object_size_bytes").record(size_bytes as f64);
     }
 }
 
@@ -367,15 +367,15 @@ pub fn record_get_object(duration_ms: f64, size_bytes: i64) {
 /// * `zero_copy_enabled` - Whether zero-copy was enabled for this operation
 #[inline(always)]
 pub fn record_put_object(duration_ms: f64, size_bytes: i64, zero_copy_enabled: bool) {
-    counter!("rustfs.s3.put_object.total").increment(1);
-    histogram!("rustfs.s3.put_object.duration.ms").record(duration_ms);
+    counter!("rustfs_s3_put_object_total").increment(1);
+    histogram!("rustfs_s3_put_object_duration_ms").record(duration_ms);
 
     if size_bytes > 0 {
-        histogram!("rustfs.s3.put_object.size.bytes").record(size_bytes as f64);
+        histogram!("rustfs_s3_put_object_size_bytes").record(size_bytes as f64);
     }
 
     if zero_copy_enabled {
-        counter!("rustfs.s3.put_object.zero_copy.enabled.total").increment(1);
+        counter!("rustfs_s3_put_object_zero_copy_enabled_total").increment(1);
     }
 }
 
@@ -388,12 +388,12 @@ pub fn record_put_object(duration_ms: f64, size_bytes: i64, zero_copy_enabled: b
 /// * `is_truncated` - Whether the response was truncated
 #[inline(always)]
 pub fn record_list_objects(duration_ms: f64, objects_count: u64, is_truncated: bool) {
-    counter!("rustfs.s3.list_objects.total").increment(1);
-    histogram!("rustfs.s3.list_objects.duration.ms").record(duration_ms);
-    histogram!("rustfs.s3.list_objects.count").record(objects_count as f64);
+    counter!("rustfs_s3_list_objects_total").increment(1);
+    histogram!("rustfs_s3_list_objects_duration_ms").record(duration_ms);
+    histogram!("rustfs_s3_list_objects_count").record(objects_count as f64);
 
     if is_truncated {
-        counter!("rustfs.s3.list_objects.truncated.total").increment(1);
+        counter!("rustfs_s3_list_objects_truncated_total").increment(1);
     }
 }
 
@@ -405,11 +405,11 @@ pub fn record_list_objects(duration_ms: f64, objects_count: u64, is_truncated: b
 /// * `version_deleted` - Whether a specific version was deleted
 #[inline(always)]
 pub fn record_delete_object(duration_ms: f64, version_deleted: bool) {
-    counter!("rustfs.s3.delete_object.total").increment(1);
-    histogram!("rustfs.s3.delete_object.duration.ms").record(duration_ms);
+    counter!("rustfs_s3_delete_object_total").increment(1);
+    histogram!("rustfs_s3_delete_object_duration_ms").record(duration_ms);
 
     if version_deleted {
-        counter!("rustfs.s3.delete_object.version.total").increment(1);
+        counter!("rustfs_s3_delete_object_version_total").increment(1);
     }
 }
 
@@ -427,18 +427,18 @@ pub fn record_delete_object(duration_ms: f64, version_deleted: bool) {
 /// * `concurrent_requests` - Number of concurrent requests
 #[inline(always)]
 pub fn record_io_strategy(storage_media: &str, access_pattern: &str, buffer_size: usize, concurrent_requests: u64) {
-    counter!("rustfs.io.strategy.total",
+    counter!("rustfs_io_strategy_total",
         "storage_media" => storage_media.to_string(),
         "access_pattern" => access_pattern.to_string(),
     )
     .increment(1);
 
-    gauge!("rustfs.io.buffer.size.bytes",
+    gauge!("rustfs_io_buffer_size_bytes",
         "storage_media" => storage_media.to_string(),
     )
     .set(buffer_size as f64);
 
-    gauge!("rustfs.io.concurrent.requests").set(concurrent_requests as f64);
+    gauge!("rustfs_io_concurrent_requests").set(concurrent_requests as f64);
 }
 
 /// Record disk permit wait time (load tracking).
@@ -448,7 +448,7 @@ pub fn record_io_strategy(storage_media: &str, access_pattern: &str, buffer_size
 /// * `duration_ms` - Time spent waiting for disk permit
 #[inline(always)]
 pub fn record_permit_wait(duration_ms: f64) {
-    histogram!("rustfs.io.permit.wait.duration.ms").record(duration_ms);
+    histogram!("rustfs_io_permit_wait_duration_ms").record(duration_ms);
 }
 
 /// Record I/O load level.
@@ -459,12 +459,12 @@ pub fn record_permit_wait(duration_ms: f64) {
 /// * `concurrent_requests` - Number of concurrent requests
 #[inline(always)]
 pub fn record_io_load_level(load_level: &str, concurrent_requests: u64) {
-    counter!("rustfs.io.load.level",
+    counter!("rustfs_io_load_level",
         "level" => load_level.to_string(),
     )
     .increment(1);
 
-    gauge!("rustfs.io.concurrent.requests").set(concurrent_requests as f64);
+    gauge!("rustfs_io_concurrent_requests").set(concurrent_requests as f64);
 }
 
 /// Record cache size and entry count.
@@ -476,12 +476,12 @@ pub fn record_io_load_level(load_level: &str, concurrent_requests: u64) {
 /// * `entries` - Number of entries in the cache
 #[inline(always)]
 pub fn record_cache_size(tier: &str, size_bytes: usize, entries: u64) {
-    gauge!("rustfs.cache.size.bytes",
+    gauge!("rustfs_cache_size_bytes",
         "tier" => tier.to_string(),
     )
     .set(size_bytes as f64);
 
-    gauge!("rustfs.cache.entries",
+    gauge!("rustfs_cache_entries",
         "tier" => tier.to_string(),
     )
     .set(entries as f64);
@@ -499,13 +499,11 @@ pub fn record_cache_size(tier: &str, size_bytes: usize, entries: u64) {
 /// * `tier` - Bandwidth tier ("low", "medium", "high", "unknown")
 #[inline(always)]
 pub fn record_bandwidth(bytes_per_second: u64, tier: &str) {
-    gauge!("rustfs.bandwidth.current.bps").set(bytes_per_second as f64);
-    gauge!("rustfs.bandwidth.current.bps",
-        "tier" => tier.to_string(),
-    )
-    .set(bytes_per_second as f64);
+    let tier_label = if tier.is_empty() { "unknown" } else { tier };
+    gauge!("rustfs_bandwidth_current_bps", "tier" => "all").set(bytes_per_second as f64);
+    gauge!("rustfs_bandwidth_current_bps", "tier" => tier_label.to_string()).set(bytes_per_second as f64);
 
-    histogram!("rustfs.bandwidth.observed.bps").record(bytes_per_second as f64);
+    histogram!("rustfs_bandwidth_observed_bps").record(bytes_per_second as f64);
 }
 
 /// Record data transfer for bandwidth calculation.
@@ -516,12 +514,12 @@ pub fn record_bandwidth(bytes_per_second: u64, tier: &str) {
 /// * `duration_ms` - Duration of the transfer in milliseconds
 #[inline(always)]
 pub fn record_data_transfer(bytes: u64, duration_ms: f64) {
-    counter!("rustfs.io.transfer.bytes").increment(bytes);
-    histogram!("rustfs.io.transfer.duration.ms").record(duration_ms);
+    counter!("rustfs_io_transfer_bytes_total").increment(bytes);
+    histogram!("rustfs_io_transfer_duration_ms").record(duration_ms);
 
     if duration_ms > 0.0 {
         let bps = (bytes as f64 * 1000.0) / duration_ms;
-        histogram!("rustfs.io.transfer.bandwidth.bps").record(bps);
+        histogram!("rustfs_io_transfer_bandwidth_bps").record(bps);
     }
 }
 
@@ -537,12 +535,12 @@ pub fn record_data_transfer(bytes: u64, duration_ms: f64) {
 /// * `total_bytes` - Total memory in bytes
 #[inline(always)]
 pub fn record_memory_usage(used_bytes: u64, total_bytes: u64) {
-    gauge!("rustfs.memory.used.bytes").set(used_bytes as f64);
-    gauge!("rustfs.memory.total.bytes").set(total_bytes as f64);
+    gauge!("rustfs_memory_used_bytes").set(used_bytes as f64);
+    gauge!("rustfs_memory_total_bytes").set(total_bytes as f64);
 
     if total_bytes > 0 {
         let usage_percent = (used_bytes as f64 / total_bytes as f64) * 100.0;
-        gauge!("rustfs.memory.usage.percent").set(usage_percent);
+        gauge!("rustfs_memory_usage_percent").set(usage_percent);
     }
 }
 
@@ -553,7 +551,7 @@ pub fn record_memory_usage(used_bytes: u64, total_bytes: u64) {
 /// * `percent` - CPU usage percentage (0.0 - 100.0)
 #[inline(always)]
 pub fn record_cpu_usage(percent: f64) {
-    gauge!("rustfs.cpu.usage.percent").set(percent);
+    gauge!("rustfs_cpu_usage_percent").set(percent);
 }
 
 /// Record disk I/O statistics.
@@ -566,13 +564,10 @@ pub fn record_cpu_usage(percent: f64) {
 /// * `write_ops` - Number of write operations
 #[inline(always)]
 pub fn record_disk_io(read_bytes: u64, write_bytes: u64, read_ops: u64, write_ops: u64) {
-    counter!("rustfs.disk.read.bytes").increment(read_bytes);
-    counter!("rustfs.disk.write.bytes").increment(write_bytes);
-    counter!("rustfs.disk.read.ops").increment(read_ops);
-    counter!("rustfs.disk.write.ops").increment(write_ops);
-
-    gauge!("rustfs.disk.read.bytes_total").set(read_bytes as f64);
-    gauge!("rustfs.disk.write.bytes_total").set(write_bytes as f64);
+    counter!("rustfs_disk_read_bytes_total").increment(read_bytes);
+    counter!("rustfs_disk_write_bytes_total").increment(write_bytes);
+    counter!("rustfs_disk_read_ops_total").increment(read_ops);
+    counter!("rustfs_disk_write_ops_total").increment(write_ops);
 }
 
 // ============================================================================
@@ -587,7 +582,7 @@ pub fn record_disk_io(read_bytes: u64, write_bytes: u64, read_ops: u64, write_op
 /// * `error_type` - Error type (e.g., "timeout", "disk_error", "network")
 #[inline(always)]
 pub fn record_error(operation: &str, error_type: &str) {
-    counter!("rustfs.errors.total",
+    counter!("rustfs_errors_total",
         "operation" => operation.to_string(),
         "type" => error_type.to_string(),
     )
@@ -602,12 +597,12 @@ pub fn record_error(operation: &str, error_type: &str) {
 /// * `duration_ms` - Duration before timeout
 #[inline(always)]
 pub fn record_timeout(operation: &str, duration_ms: f64) {
-    counter!("rustfs.timeouts.total",
+    counter!("rustfs_timeouts_total",
         "operation" => operation.to_string(),
     )
     .increment(1);
 
-    histogram!("rustfs.timeouts.duration.ms",
+    histogram!("rustfs_timeouts_duration_ms",
         "operation" => operation.to_string(),
     )
     .record(duration_ms);
@@ -621,12 +616,12 @@ pub fn record_timeout(operation: &str, duration_ms: f64) {
 /// * `attempt_number` - Attempt number (1-based)
 #[inline(always)]
 pub fn record_retry(operation: &str, attempt_number: u32) {
-    counter!("rustfs.retries.total",
+    counter!("rustfs_retries_total",
         "operation" => operation.to_string(),
     )
     .increment(1);
 
-    histogram!("rustfs.retries.attempt",
+    histogram!("rustfs_retries_attempt",
         "operation" => operation.to_string(),
     )
     .record(attempt_number as f64);
@@ -643,7 +638,7 @@ pub fn record_retry(operation: &str, attempt_number: u32) {
 /// * `latency_ms` - I/O latency in milliseconds
 #[inline(always)]
 pub fn record_io_latency(latency_ms: f64) {
-    histogram!("rustfs.io.latency.ms").record(latency_ms);
+    histogram!("rustfs_io_latency_ms").record(latency_ms);
 }
 
 /// Record I/O latency P95 in milliseconds.
@@ -653,7 +648,7 @@ pub fn record_io_latency(latency_ms: f64) {
 /// * `latency_ms` - P95 I/O latency in milliseconds
 #[inline(always)]
 pub fn record_io_latency_p95(latency_ms: f64) {
-    gauge!("rustfs.io.latency.p95.ms").set(latency_ms);
+    gauge!("rustfs_io_latency_p95_ms").set(latency_ms);
 }
 
 /// Record I/O latency P99 in milliseconds.
@@ -663,7 +658,7 @@ pub fn record_io_latency_p95(latency_ms: f64) {
 /// * `latency_ms` - P99 I/O latency in milliseconds
 #[inline(always)]
 pub fn record_io_latency_p99(latency_ms: f64) {
-    gauge!("rustfs.io.latency.p99.ms").set(latency_ms);
+    gauge!("rustfs_io_latency_p99_ms").set(latency_ms);
 }
 
 #[cfg(test)]

crates/io-metrics/src/lock_metrics.rs

@@ -20,7 +20,7 @@ use std::time::Duration;
 #[inline(always)]
 pub fn record_lock_optimization_enabled(enabled: bool) {
     use metrics::gauge;
-    gauge!("rustfs.lock.optimization.enabled").set(if enabled { 1.0 } else { 0.0 });
+    gauge!("rustfs_lock_optimization_enabled").set(if enabled { 1.0 } else { 0.0 });
 }
 
 /// Record spin attempt.
@@ -28,9 +28,9 @@ pub fn record_lock_optimization_enabled(enabled: bool) {
 pub fn record_spin_attempt(success: bool) {
     use metrics::counter;
     if success {
-        counter!("rustfs.lock.spin.successes").increment(1);
+        counter!("rustfs_lock_spin_successes").increment(1);
     } else {
-        counter!("rustfs.lock.spin.failures").increment(1);
+        counter!("rustfs_lock_spin_failures").increment(1);
     }
 }
 
@@ -38,28 +38,28 @@ pub fn record_spin_attempt(success: bool) {
 #[inline(always)]
 pub fn record_spin_count_change(new_count: usize) {
     use metrics::gauge;
-    gauge!("rustfs.lock.spin.count").set(new_count as f64);
+    gauge!("rustfs_lock_spin_count").set(new_count as f64);
 }
 
 /// Record lock hold time.
 #[inline(always)]
 pub fn record_lock_hold_time(hold_time: Duration) {
     use metrics::histogram;
-    histogram!("rustfs.lock.hold_time.secs").record(hold_time.as_secs_f64());
+    histogram!("rustfs_lock_hold_time_secs").record(hold_time.as_secs_f64());
 }
 
 /// Record early release.
 #[inline(always)]
 pub fn record_early_release() {
     use metrics::counter;
-    counter!("rustfs.lock.early_releases").increment(1);
+    counter!("rustfs_lock_early_releases").increment(1);
 }
 
 /// Record contention event.
 #[inline(always)]
 pub fn record_contention_event() {
     use metrics::counter;
-    counter!("rustfs.lock.contentions").increment(1);
+    counter!("rustfs_lock_contentions").increment(1);
 }
 
 /// Lock statistics summary.

crates/io-metrics/src/metric_names.rs

@@ -49,6 +49,6 @@ pub mod zero_copy {
     /// Throughput in MB/s
     pub const THROUGHPUT_MBPS: &str = "rustfs_zero_copy_throughput_mbps";
 
-    /// Memory saved by zero-copy in bytes
-    pub const MEMORY_SAVED_BYTES: &str = "rustfs_zero_copy_memory_saved_bytes";
+    /// Current memory saved estimate by zero-copy in bytes
+    pub const MEMORY_SAVED_BYTES: &str = "rustfs_zero_copy_memory_saved_bytes_current";
 }

crates/io-metrics/src/timeout_metrics.rs

@@ -34,23 +34,23 @@ pub fn record_operation_duration(operation: &str, duration: Duration) {
 #[inline(always)]
 pub fn record_dynamic_timeout(size_bytes: u64, timeout: Duration) {
     use metrics::{gauge, histogram};
-    gauge!("rustfs.timeout.dynamic.size").set(size_bytes as f64);
-    gauge!("rustfs.timeout.dynamic.secs").set(timeout.as_secs_f64());
-    histogram!("rustfs.timeout.dynamic.size.histogram").record(size_bytes as f64);
+    gauge!("rustfs_timeout_dynamic_size").set(size_bytes as f64);
+    gauge!("rustfs_timeout_dynamic_secs").set(timeout.as_secs_f64());
+    histogram!("rustfs_timeout_dynamic_size_histogram").record(size_bytes as f64);
 }
 
 /// Record operation progress.
 #[inline(always)]
 pub fn record_operation_progress(operation: &str, percent: f64) {
     use metrics::gauge;
-    gauge!("rustfs.operation.progress", "operation" => operation.to_string()).set(percent);
+    gauge!("rustfs_operation_progress", "operation" => operation.to_string()).set(percent);
 }
 
 /// Record stalled operation.
 #[inline(always)]
 pub fn record_stalled_operation(operation: &str) {
     use metrics::counter;
-    counter!("rustfs.operation.stalled", "operation" => operation.to_string()).increment(1);
+    counter!("rustfs_operation_stalled", "operation" => operation.to_string()).increment(1);
 }
 
 /// Record operation completion.
@@ -58,7 +58,7 @@ pub fn record_stalled_operation(operation: &str) {
 pub fn record_operation_completion(operation: &str, success: bool) {
     use metrics::counter;
     let status = if success { "success" } else { "failure" };
-    counter!("rustfs.operation.completions", "operation" => operation.to_string(), "status" => status).increment(1);
+    counter!("rustfs_operation_completions", "operation" => operation.to_string(), "status" => status).increment(1);
 }
 
 /// Timeout statistics summary.

crates/obs/Cargo.toml

@@ -34,11 +34,14 @@ workspace = true
 
 [dependencies]
 rustfs-audit = { workspace = true }
+rustfs-common = { workspace = true }
 rustfs-config = { workspace = true, features = ["constants", "observability"] }
 rustfs-ecstore = { workspace = true }
+rustfs-iam = { workspace = true }
 rustfs-io-metrics = { workspace = true }
 rustfs-notify = { workspace = true }
 rustfs-utils = { workspace = true, features = ["ip"] }
+chrono = { workspace = true }
 flate2 = { workspace = true }
 glob = { workspace = true }
 jiff = { workspace = true }

crates/obs/README.md

@@ -171,16 +171,16 @@ The log rotation and cleanup pipeline emits these metrics (via the `metrics` fac
 
 | Metric | Type | Description |
 |---|---|---|
-| `rustfs.log_cleaner.deleted_files_total` | counter | Number of files deleted per cleanup pass |
-| `rustfs.log_cleaner.freed_bytes_total` | counter | Bytes reclaimed by deletion |
-| `rustfs.log_cleaner.compress_duration_seconds` | histogram | Compression stage duration |
-| `rustfs.log_cleaner.steal_success_rate` | gauge | Work-stealing success ratio in parallel mode |
-| `rustfs.log_cleaner.runs_total` | counter | Successful cleanup loop runs |
-| `rustfs.log_cleaner.run_failures_total` | counter | Failed or panicked cleanup loop runs |
-| `rustfs.log_cleaner.rotation_total` | counter | Successful file rotations |
-| `rustfs.log_cleaner.rotation_failures_total` | counter | Failed file rotations |
-| `rustfs.log_cleaner.rotation_duration_seconds` | histogram | Rotation latency |
-| `rustfs.log_cleaner.active_file_size_bytes` | gauge | Current active log file size |
+| `rustfs_log_cleaner_deleted_files_total` | counter | Number of files deleted per cleanup pass |
+| `rustfs_log_cleaner_freed_bytes_total` | counter | Bytes reclaimed by deletion |
+| `rustfs_log_cleaner_compress_duration_seconds` | histogram | Compression stage duration |
+| `rustfs_log_cleaner_steal_success_rate` | gauge | Work-stealing success ratio in parallel mode |
+| `rustfs_log_cleaner_runs_total` | counter | Successful cleanup loop runs |
+| `rustfs_log_cleaner_run_failures_total` | counter | Failed or panicked cleanup loop runs |
+| `rustfs_log_cleaner_rotation_total` | counter | Successful file rotations |
+| `rustfs_log_cleaner_rotation_failures_total` | counter | Failed file rotations |
+| `rustfs_log_cleaner_rotation_duration_seconds` | histogram | Rotation latency |
+| `rustfs_log_cleaner_active_file_size_bytes` | gauge | Current active log file size |
 
 These metrics cover compression, cleanup, and file rotation end-to-end.
 
@@ -195,8 +195,8 @@ These metrics cover compression, cleanup, and file rotation end-to-end.
 ### Grafana Dashboard JSON Draft (Ready to Import)
 
 > Save this as `rustfs-log-cleaner-dashboard.json`, then import from Grafana UI.
-> For Prometheus datasources, metric names are usually normalized to underscores,
-> so `rustfs.log_cleaner.deleted_files_total` becomes `rustfs_log_cleaner_deleted_files_total`.
+> The canonical metric names use underscore notation, for example
+> `rustfs_log_cleaner_deleted_files_total`.
 >
 > The same panels are now checked in at:
 > `.docker/observability/grafana/dashboards/rustfs.json`

crates/obs/src/cleaner/README.md

@@ -58,14 +58,14 @@ This strategy keeps local cache affinity while still balancing stragglers.
 
 The cleaner emits tracing events and runtime metrics:
 
-- `rustfs.log_cleaner.deleted_files_total` (counter)
-- `rustfs.log_cleaner.freed_bytes_total` (counter)
-- `rustfs.log_cleaner.compress_duration_seconds` (histogram)
-- `rustfs.log_cleaner.steal_success_rate` (gauge)
-- `rustfs.log_cleaner.rotation_total` (counter)
-- `rustfs.log_cleaner.rotation_failures_total` (counter)
-- `rustfs.log_cleaner.rotation_duration_seconds` (histogram)
-- `rustfs.log_cleaner.active_file_size_bytes` (gauge)
+- `rustfs_log_cleaner_deleted_files_total` (counter)
+- `rustfs_log_cleaner_freed_bytes_total` (counter)
+- `rustfs_log_cleaner_compress_duration_seconds` (histogram)
+- `rustfs_log_cleaner_steal_success_rate` (gauge)
+- `rustfs_log_cleaner_rotation_total` (counter)
+- `rustfs_log_cleaner_rotation_failures_total` (counter)
+- `rustfs_log_cleaner_rotation_duration_seconds` (histogram)
+- `rustfs_log_cleaner_active_file_size_bytes` (gauge)
 
 These values can be wired into dashboards and alert rules for cleanup health.
 
@@ -127,4 +127,3 @@ let _ = cleaner.cleanup();
 - Prefer `FileMatchMode::Suffix` when rotations prepend timestamps to the filename.
 - Prefer `FileMatchMode::Prefix` when rotations append counters or timestamps after a stable base name.
 - Keep `parallel_workers` modest when `zstd_workers` is greater than `1`, because each compression task may already use internal codec threads.
-

crates/obs/src/metrics/collectors/cluster_iam.rs

@@ -25,6 +25,10 @@ use crate::metrics::schema::cluster_iam::*;
 /// IAM statistics.
 #[derive(Debug, Clone, Default)]
 pub struct IamStats {
+    /// Last successful IAM data sync duration in milliseconds
+    pub last_sync_duration_millis: u64,
+    /// Failed requests count in the last full minute
+    pub plugin_authn_service_failed_requests_minute: u64,
     /// Time in seconds since last failed authn service request
     pub plugin_authn_service_last_fail_seconds: u64,
     /// Time in seconds since last successful authn service request
@@ -48,6 +52,11 @@ pub struct IamStats {
 /// Returns a vector of Prometheus metrics for IAM statistics.
 pub fn collect_iam_metrics(stats: &IamStats) -> Vec<PrometheusMetric> {
     vec![
+        PrometheusMetric::from_descriptor(&LAST_SYNC_DURATION_MILLIS_MD, stats.last_sync_duration_millis as f64),
+        PrometheusMetric::from_descriptor(
+            &PLUGIN_AUTHN_SERVICE_FAILED_REQUESTS_MINUTE_MD,
+            stats.plugin_authn_service_failed_requests_minute as f64,
+        ),
         PrometheusMetric::from_descriptor(
             &PLUGIN_AUTHN_SERVICE_LAST_FAIL_SECONDS_MD,
             stats.plugin_authn_service_last_fail_seconds as f64,
@@ -82,6 +91,8 @@ mod tests {
     #[test]
     fn test_collect_iam_metrics() {
         let stats = IamStats {
+            last_sync_duration_millis: 250,
+            plugin_authn_service_failed_requests_minute: 7,
             plugin_authn_service_last_fail_seconds: 3600,
             plugin_authn_service_last_succ_seconds: 10,
             plugin_authn_service_succ_avg_rtt_ms_minute: 50,
@@ -95,7 +106,7 @@ mod tests {
         let metrics = collect_iam_metrics(&stats);
         report_metrics(&metrics);
 
-        assert_eq!(metrics.len(), 8);
+        assert_eq!(metrics.len(), 10);
 
         let sync_successes_name = SYNC_SUCCESSES_MD.get_full_metric_name();
         let sync_successes = metrics.iter().find(|m| m.name == sync_successes_name);
@@ -108,7 +119,7 @@ mod tests {
         let stats = IamStats::default();
         let metrics = collect_iam_metrics(&stats);
 
-        assert_eq!(metrics.len(), 8);
+        assert_eq!(metrics.len(), 10);
         for metric in &metrics {
             assert_eq!(metric.value, 0.0);
             assert!(metric.labels.is_empty());

crates/obs/src/metrics/scheduler.rs

@@ -36,13 +36,20 @@ use crate::metrics::collectors::{
     collect_bucket_metrics,
     collect_bucket_replication_bandwidth_metrics,
     collect_bucket_replication_metrics,
+    collect_bucket_usage_metrics,
+    collect_cluster_config_metrics,
     collect_cluster_health_metrics,
     collect_cluster_metrics,
+    collect_cluster_usage_metrics,
     collect_cpu_metrics,
     collect_drive_count_metrics,
     collect_drive_detailed_metrics,
+    collect_erasure_set_metrics,
     collect_host_network_metrics,
+    collect_iam_metrics,
+    collect_ilm_metrics,
     collect_memory_metrics,
+    collect_network_metrics,
     collect_node_metrics,
     collect_notification_metrics,
     collect_notification_target_metrics,
@@ -53,6 +60,7 @@ use crate::metrics::collectors::{
     collect_process_metrics,
     collect_replication_metrics,
     collect_resource_metrics,
+    collect_scanner_metrics,
 };
 use crate::metrics::config::{
     DEFAULT_AUDIT_METRICS_INTERVAL, DEFAULT_BUCKET_METRICS_INTERVAL, DEFAULT_BUCKET_REPLICATION_BANDWIDTH_METRICS_INTERVAL,
@@ -64,8 +72,10 @@ use crate::metrics::config::{
 use crate::metrics::report::{PrometheusMetric, report_metrics};
 use crate::metrics::stats_collector::{
     ProcessMetricBundle, collect_bucket_replication_bandwidth_stats, collect_bucket_replication_detail_stats,
-    collect_bucket_stats, collect_cluster_and_health_stats, collect_disk_and_system_drive_stats, collect_host_network_stats,
-    collect_process_metric_bundle, collect_replication_stats, collect_system_cpu_and_memory_stats_with,
+    collect_bucket_stats, collect_cluster_and_health_stats, collect_cluster_config_stats, collect_cluster_usage_metric_stats,
+    collect_disk_and_system_drive_stats, collect_erasure_set_stats, collect_host_network_stats, collect_iam_stats,
+    collect_ilm_metric_stats, collect_internode_network_stats, collect_process_metric_bundle, collect_replication_stats,
+    collect_scanner_metric_stats, collect_system_cpu_and_memory_stats_with,
 };
 use rustfs_audit::audit_target_metrics;
 use rustfs_notify::{notification_metrics_snapshot, notification_target_metrics};
@@ -176,6 +186,46 @@ pub fn init_metrics_runtime(token: CancellationToken) {
         }
     });
 
+    // Spawn task for supplementary cluster metrics that are defined in schema/collector
+    // but filled by later task-specific runtime sources.
+    let token_clone = token.clone();
+    tokio::spawn(async move {
+        let mut interval = tokio::time::interval(cluster_interval);
+        loop {
+            tokio::select! {
+                _ = interval.tick() => {
+                    let mut metrics = Vec::new();
+
+                    if let Some(stats) = collect_cluster_config_stats().await {
+                        metrics.extend(collect_cluster_config_metrics(&stats));
+                    }
+
+                    let erasure_sets = collect_erasure_set_stats().await;
+                    if !erasure_sets.is_empty() {
+                        metrics.extend(collect_erasure_set_metrics(&erasure_sets));
+                    }
+
+                    if let Some(stats) = collect_iam_stats().await {
+                        metrics.extend(collect_iam_metrics(&stats));
+                    }
+
+                    if let Some((cluster_usage, bucket_usage)) = collect_cluster_usage_metric_stats().await {
+                        metrics.extend(collect_cluster_usage_metrics(&cluster_usage));
+                        metrics.extend(collect_bucket_usage_metrics(&bucket_usage));
+                    }
+
+                    if !metrics.is_empty() {
+                        report_metrics(&metrics);
+                    }
+                }
+                _ = token_clone.cancelled() => {
+                    warn!("Metrics collection for supplementary cluster stats cancelled.");
+                    return;
+                }
+            }
+        }
+    });
+
     // Spawn task for bucket metrics
     let token_clone = token.clone();
     tokio::spawn(async move {
@@ -302,6 +352,35 @@ pub fn init_metrics_runtime(token: CancellationToken) {
         }
     });
 
+    // Spawn task for background workflow metrics such as ILM and scanner.
+    let token_clone = token.clone();
+    tokio::spawn(async move {
+        let mut interval = tokio::time::interval(cluster_interval);
+        loop {
+            tokio::select! {
+                _ = interval.tick() => {
+                    let mut metrics = Vec::new();
+
+                    if let Some(stats) = collect_ilm_metric_stats().await {
+                        metrics.extend(collect_ilm_metrics(&stats));
+                    }
+
+                    if let Some(stats) = collect_scanner_metric_stats().await {
+                        metrics.extend(collect_scanner_metrics(&stats));
+                    }
+
+                    if !metrics.is_empty() {
+                        report_metrics(&metrics);
+                    }
+                }
+                _ = token_clone.cancelled() => {
+                    warn!("Metrics collection for background workflow stats cancelled.");
+                    return;
+                }
+            }
+        }
+    });
+
     // Spawn task for system monitoring metrics (migrated from rustfs-obs::system)
     let system_interval = get_env_opt_u64(ENV_SYSTEM_METRICS_INTERVAL)
         .or_else(|| get_env_opt_u64(LEGACY_SYSTEM_METRICS_INTERVAL).map(|ms| ms / 1000)) // Convert ms to seconds
@@ -310,7 +389,7 @@ pub fn init_metrics_runtime(token: CancellationToken) {
         .map(Duration::from_secs)
         .unwrap_or(DEFAULT_SYSTEM_METRICS_INTERVAL);
 
-    let token_clone = token;
+    let token_clone = token.clone();
     tokio::spawn(async move {
         let labels = current_process_metric_labels();
         let mut host_system = System::new_all();
@@ -378,6 +457,28 @@ pub fn init_metrics_runtime(token: CancellationToken) {
             }
         }
     });
+
+    // Spawn task for internode/system network metrics.
+    let token_clone = token;
+    tokio::spawn(async move {
+        let mut interval = tokio::time::interval(system_interval);
+        loop {
+            tokio::select! {
+                _ = interval.tick() => {
+                    if let Some(stats) = collect_internode_network_stats() {
+                        let metrics = collect_network_metrics(&stats);
+                        if !metrics.is_empty() {
+                            report_metrics(&metrics);
+                        }
+                    }
+                }
+                _ = token_clone.cancelled() => {
+                    warn!("Metrics collection for internode network stats cancelled.");
+                    return;
+                }
+            }
+        }
+    });
 }
 
 /// Backward-compatible alias kept during migration.

crates/obs/src/metrics/stats_collector.rs

@@ -21,10 +21,14 @@
 //! and convert them to the Stats structs used by collectors.
 
 use crate::metrics::collectors::{
-    BucketReplicationBandwidthStats, BucketReplicationStats, BucketReplicationTargetStats, BucketStats, ClusterHealthStats,
-    ClusterStats, CpuStats, DiskStats, DriveCountStats, DriveDetailedStats, HostNetworkStats, MemoryStats, ProcessStats,
-    ProcessStatusType, ReplicationStats, ResourceStats,
+    BucketReplicationBandwidthStats, BucketReplicationStats, BucketReplicationTargetStats, BucketStats, BucketUsageStats,
+    ClusterConfigStats, ClusterHealthStats, ClusterStats, ClusterUsageStats, CpuStats, DiskStats, DriveCountStats,
+    DriveDetailedStats, ErasureSetStats, HostNetworkStats, IamStats, IlmStats, MemoryStats, NetworkStats, ProcessStats,
+    ProcessStatusType, ReplicationStats, ResourceStats, ScannerStats,
 };
+use chrono::Utc;
+use rustfs_common::{internode_metrics::global_internode_metrics, metrics::global_metrics};
+use rustfs_ecstore::bucket::lifecycle::bucket_lifecycle_ops::{GLOBAL_ExpiryState, GLOBAL_TransitionState};
 use rustfs_ecstore::bucket::metadata_sys::get_quota_config;
 use rustfs_ecstore::bucket::replication::GLOBAL_REPLICATION_STATS;
 use rustfs_ecstore::data_usage::load_data_usage_from_backend;
@@ -32,7 +36,9 @@ use rustfs_ecstore::global::get_global_bucket_monitor;
 use rustfs_ecstore::pools::{get_total_usable_capacity, get_total_usable_capacity_free};
 use rustfs_ecstore::store_api::{BucketOperations, BucketOptions};
 use rustfs_ecstore::{StorageAPI, new_object_layer_fn};
+use rustfs_iam::{get_global_iam_sys, oidc::oidc_plugin_authn_metrics_snapshot};
 use rustfs_io_metrics::{ProcessStatusSnapshot, snapshot_process_resource_and_system};
+use std::collections::HashMap;
 use std::time::Duration;
 use sysinfo::{Networks, System};
 use tracing::{instrument, warn};
@@ -42,6 +48,15 @@ const DRIVE_STATE_ONLINE: &str = "online";
 const DRIVE_STATE_UNFORMATTED: &str = "unformatted";
 const DRIVE_RUNTIME_STATE_RETURNING: &str = "returning";
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+struct ErasureSetQuorumShape {
+    data_shards: u32,
+    read_quorum: u32,
+    write_quorum: u32,
+    read_tolerance: u32,
+    write_tolerance: u32,
+}
+
 fn disk_is_online_for_metrics(state: &str, runtime_state: Option<&str>) -> bool {
     let state_is_acceptable = state.eq_ignore_ascii_case(DRIVE_STATE_OK)
         || state.eq_ignore_ascii_case(DRIVE_STATE_ONLINE)
@@ -56,6 +71,30 @@ fn disk_is_online_for_metrics(state: &str, runtime_state: Option<&str>) -> bool
     state_is_acceptable
 }
 
+fn derive_erasure_set_quorum_shape(set_drive_count: usize, parity: usize) -> ErasureSetQuorumShape {
+    let data_shards = set_drive_count.saturating_sub(parity);
+    let read_quorum = data_shards.max(1);
+    let mut write_quorum = read_quorum;
+    if data_shards == parity {
+        write_quorum += 1;
+    }
+
+    ErasureSetQuorumShape {
+        data_shards: data_shards as u32,
+        read_quorum: read_quorum as u32,
+        write_quorum: write_quorum as u32,
+        read_tolerance: parity as u32,
+        write_tolerance: set_drive_count.saturating_sub(write_quorum) as u32,
+    }
+}
+
+fn apply_erasure_set_health(entry: &mut ErasureSetStats) {
+    let online = entry.online_drives_count;
+    entry.read_health = u8::from(online >= entry.read_quorum);
+    entry.write_health = u8::from(online >= entry.write_quorum);
+    entry.health = u8::from(entry.write_health == 1);
+}
+
 #[derive(Debug, Clone, Default)]
 pub struct ProcessMetricBundle {
     pub resource: ResourceStats,
@@ -285,14 +324,12 @@ pub async fn collect_bucket_replication_detail_stats() -> Vec<BucketReplicationS
             proxied_get_requests_failures: proxy.get_failed.max(0) as u64,
             proxied_head_requests_total: proxy.head_total.max(0) as u64,
             proxied_head_requests_failures: proxy.head_failed.max(0) as u64,
-            // Proxy cache currently tracks generic PutObject requests, not tagging-specific APIs.
-            // Keep tagging counters zero until PutObjectTagging stats are tracked separately.
-            proxied_put_tagging_requests_total: 0,
-            proxied_put_tagging_requests_failures: 0,
-            proxied_get_tagging_requests_total: 0,
-            proxied_get_tagging_requests_failures: 0,
-            proxied_delete_tagging_requests_total: 0,
-            proxied_delete_tagging_requests_failures: 0,
+            proxied_put_tagging_requests_total: proxy.put_tag_total.max(0) as u64,
+            proxied_put_tagging_requests_failures: proxy.put_tag_failed.max(0) as u64,
+            proxied_get_tagging_requests_total: proxy.get_tag_total.max(0) as u64,
+            proxied_get_tagging_requests_failures: proxy.get_tag_failed.max(0) as u64,
+            proxied_delete_tagging_requests_total: proxy.delete_tag_total.max(0) as u64,
+            proxied_delete_tagging_requests_failures: proxy.delete_tag_failed.max(0) as u64,
             targets,
         });
     }
@@ -585,6 +622,226 @@ pub fn collect_host_network_stats() -> HostNetworkStats {
     }
 }
 
+/// Collect internode network metrics from the global internode metrics snapshot.
+///
+/// The returned values come directly from `global_internode_metrics().snapshot()`
+/// and currently include only the counters and dial timing data tracked by the
+/// internode metrics runtime.
+pub fn collect_internode_network_stats() -> Option<NetworkStats> {
+    let snapshot = global_internode_metrics().snapshot();
+
+    Some(NetworkStats {
+        internode_errors_total: snapshot.errors_total,
+        internode_dial_errors_total: snapshot.dial_errors_total,
+        internode_dial_avg_time_nanos: snapshot.dial_avg_time_nanos,
+        internode_sent_bytes_total: snapshot.sent_bytes_total,
+        internode_recv_bytes_total: snapshot.recv_bytes_total,
+    })
+}
+
+/// Collect cluster config metrics from backend parity configuration.
+pub async fn collect_cluster_config_stats() -> Option<ClusterConfigStats> {
+    let store = new_object_layer_fn()?;
+    let backend = store.backend_info().await;
+
+    Some(ClusterConfigStats {
+        rrs_parity: backend.rr_sc_parity.unwrap_or_default() as u32,
+        standard_parity: backend.standard_sc_parity.unwrap_or_default() as u32,
+    })
+}
+
+/// Collect cluster erasure set metrics from storage and backend topology info.
+pub async fn collect_erasure_set_stats() -> Vec<ErasureSetStats> {
+    let Some(store) = new_object_layer_fn() else {
+        return Vec::new();
+    };
+
+    let storage_info = store.storage_info().await;
+    let backend = store.backend_info().await;
+    let mut grouped: HashMap<(usize, usize), ErasureSetStats> = HashMap::new();
+
+    for disk in &storage_info.disks {
+        let pool_idx = disk.pool_index.max(0) as usize;
+        let set_idx = disk.set_index.max(0) as usize;
+        let set_drive_count = backend.drives_per_set.get(pool_idx).copied().unwrap_or_default();
+        let parity = backend
+            .standard_sc_parities
+            .get(pool_idx)
+            .copied()
+            .or(backend.standard_sc_parity)
+            .unwrap_or(set_drive_count / 2);
+        let quorum_shape = derive_erasure_set_quorum_shape(set_drive_count, parity);
+
+        let entry = grouped.entry((pool_idx, set_idx)).or_insert_with(|| ErasureSetStats {
+            pool_id: pool_idx as u32,
+            set_id: set_idx as u32,
+            size: set_drive_count as u32,
+            parity: parity as u32,
+            data_shards: quorum_shape.data_shards,
+            read_quorum: quorum_shape.read_quorum,
+            write_quorum: quorum_shape.write_quorum,
+            online_drives_count: 0,
+            healing_drives_count: 0,
+            health: 0,
+            read_tolerance: quorum_shape.read_tolerance,
+            write_tolerance: quorum_shape.write_tolerance,
+            read_health: 0,
+            write_health: 0,
+        });
+
+        if disk_is_online_for_metrics(disk.state.as_str(), disk.runtime_state.as_deref()) {
+            entry.online_drives_count += 1;
+        }
+        if disk.healing {
+            entry.healing_drives_count += 1;
+        }
+    }
+
+    for entry in grouped.values_mut() {
+        apply_erasure_set_health(entry);
+    }
+
+    let mut stats = grouped.into_values().collect::<Vec<_>>();
+    stats.sort_by_key(|stat| (stat.pool_id, stat.set_id));
+    stats
+}
+
+pub async fn collect_iam_stats() -> Option<IamStats> {
+    let iam_sys = get_global_iam_sys()?;
+    let sync = iam_sys.sync_metrics_snapshot();
+    let oidc = oidc_plugin_authn_metrics_snapshot();
+
+    Some(IamStats {
+        last_sync_duration_millis: sync.last_sync_duration_millis,
+        plugin_authn_service_failed_requests_minute: oidc.failed_requests_minute,
+        plugin_authn_service_last_fail_seconds: oidc.last_fail_seconds,
+        plugin_authn_service_last_succ_seconds: oidc.last_succ_seconds,
+        plugin_authn_service_succ_avg_rtt_ms_minute: oidc.succ_avg_rtt_ms_minute,
+        plugin_authn_service_succ_max_rtt_ms_minute: oidc.succ_max_rtt_ms_minute,
+        plugin_authn_service_total_requests_minute: oidc.total_requests_minute,
+        since_last_sync_millis: sync.since_last_sync_millis,
+        sync_failures: sync.sync_failures,
+        sync_successes: sync.sync_successes,
+    })
+}
+
+/// Collect cluster and per-bucket usage metrics from backend usage snapshots.
+///
+/// This reads persisted usage data via `load_data_usage_from_backend()` and
+/// builds cluster totals plus per-bucket distributions from the returned
+/// histograms. It does not trigger an inline object-data rescan.
+pub async fn collect_cluster_usage_metric_stats() -> Option<(ClusterUsageStats, Vec<BucketUsageStats>)> {
+    let store = new_object_layer_fn()?;
+    let data_usage = load_data_usage_from_backend(store.clone()).await.ok()?;
+    let mut buckets = Vec::with_capacity(data_usage.buckets_usage.len());
+
+    for (bucket_name, usage) in &data_usage.buckets_usage {
+        if bucket_name.starts_with('.') {
+            continue;
+        }
+
+        let quota_bytes = match get_quota_config(bucket_name).await {
+            Ok((quota, _)) => quota.get_quota_limit().unwrap_or(0),
+            Err(_) => 0,
+        };
+
+        buckets.push(BucketUsageStats {
+            bucket: bucket_name.clone(),
+            total_bytes: usage.size,
+            objects_count: usage.objects_count,
+            versions_count: usage.versions_count,
+            delete_markers_count: usage.delete_markers_count,
+            quota_bytes,
+            object_size_distribution: usage
+                .object_size_histogram
+                .iter()
+                .map(|(range, count)| (range.clone(), *count))
+                .collect(),
+            version_count_distribution: usage
+                .object_versions_histogram
+                .iter()
+                .map(|(range, count)| (range.clone(), *count))
+                .collect(),
+        });
+    }
+
+    buckets.sort_by(|a, b| a.bucket.cmp(&b.bucket));
+
+    Some((
+        ClusterUsageStats {
+            total_bytes: data_usage.objects_total_size,
+            objects_count: data_usage.objects_total_count,
+            versions_count: data_usage.versions_total_count,
+            delete_markers_count: data_usage.delete_markers_total_count,
+            object_size_distribution: data_usage
+                .buckets_usage
+                .values()
+                .flat_map(|usage| usage.object_size_histogram.iter())
+                .fold(HashMap::<String, u64>::new(), |mut acc, (range, count)| {
+                    *acc.entry(range.clone()).or_default() += *count;
+                    acc
+                })
+                .into_iter()
+                .collect(),
+            versions_distribution: data_usage
+                .buckets_usage
+                .values()
+                .flat_map(|usage| usage.object_versions_histogram.iter())
+                .fold(HashMap::<String, u64>::new(), |mut acc, (range, count)| {
+                    *acc.entry(range.clone()).or_default() += *count;
+                    acc
+                })
+                .into_iter()
+                .collect(),
+        },
+        buckets,
+    ))
+}
+
+/// Collect ILM metrics from the current lifecycle runtime state.
+pub async fn collect_ilm_metric_stats() -> Option<IlmStats> {
+    let expiry_pending_tasks = GLOBAL_ExpiryState.read().await.pending_tasks().await as u64;
+    let transition_active_tasks = GLOBAL_TransitionState.active_tasks().max(0) as u64;
+    let transition_pending_tasks = GLOBAL_TransitionState.pending_tasks() as u64;
+    let transition_missed_immediate_tasks = GLOBAL_TransitionState.missed_immediate_tasks().max(0) as u64;
+    let metrics = global_metrics().report().await;
+    let versions_scanned = metrics.life_time_ilm.values().copied().sum();
+
+    Some(IlmStats {
+        expiry_pending_tasks,
+        transition_active_tasks,
+        transition_pending_tasks,
+        transition_missed_immediate_tasks,
+        versions_scanned,
+    })
+}
+
+/// Collect scanner metrics from a runtime source.
+///
+/// Task 5 maps scanner runtime snapshots from `global_metrics()` into the
+/// rustfs-obs scanner collector shape.
+pub async fn collect_scanner_metric_stats() -> Option<ScannerStats> {
+    let metrics = global_metrics().report().await;
+    let bucket_scans_finished = metrics.life_time_ops.get("scan_bucket_drive").copied().unwrap_or_default();
+    let directories_scanned = metrics.life_time_ops.get("scan_folder").copied().unwrap_or_default();
+    let objects_scanned = metrics.life_time_ops.get("scan_object").copied().unwrap_or_default();
+    let versions_scanned = metrics.life_time_ilm.values().copied().sum();
+    let reference_time = metrics.cycles_completed_at.last().copied().unwrap_or(metrics.current_started);
+    let last_activity_seconds = Utc::now().signed_duration_since(reference_time).num_seconds().max(0) as u64;
+
+    Some(ScannerStats {
+        bucket_scans_finished,
+        // `global_metrics()` currently tracks completed bucket-drive scans, not a
+        // separate started counter. Mirror the finished count until Task 5/Task 10
+        // expands the scanner runtime source shape.
+        bucket_scans_started: bucket_scans_finished,
+        directories_scanned,
+        objects_scanned,
+        versions_scanned,
+        last_activity_seconds,
+    })
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -598,4 +855,57 @@ mod tests {
     fn disk_is_online_for_metrics_rejects_offline_runtime_state() {
         assert!(!disk_is_online_for_metrics(DRIVE_STATE_OK, Some("offline")));
     }
+
+    #[test]
+    fn derive_erasure_set_quorum_shape_handles_standard_layout() {
+        let shape = derive_erasure_set_quorum_shape(16, 4);
+
+        assert_eq!(
+            shape,
+            ErasureSetQuorumShape {
+                data_shards: 12,
+                read_quorum: 12,
+                write_quorum: 12,
+                read_tolerance: 4,
+                write_tolerance: 4,
+            }
+        );
+    }
+
+    #[test]
+    fn derive_erasure_set_quorum_shape_handles_equal_data_and_parity() {
+        let shape = derive_erasure_set_quorum_shape(4, 2);
+
+        assert_eq!(
+            shape,
+            ErasureSetQuorumShape {
+                data_shards: 2,
+                read_quorum: 2,
+                write_quorum: 3,
+                read_tolerance: 2,
+                write_tolerance: 1,
+            }
+        );
+    }
+
+    #[test]
+    fn apply_erasure_set_health_marks_read_and_write_health_from_online_count() {
+        let mut stats = ErasureSetStats {
+            read_quorum: 3,
+            write_quorum: 4,
+            online_drives_count: 3,
+            ..Default::default()
+        };
+
+        apply_erasure_set_health(&mut stats);
+        assert_eq!(stats.read_health, 1);
+        assert_eq!(stats.write_health, 0);
+        assert_eq!(stats.health, 0);
+
+        stats.online_drives_count = 4;
+        apply_erasure_set_health(&mut stats);
+        assert_eq!(stats.read_health, 1);
+        assert_eq!(stats.write_health, 1);
+        assert_eq!(stats.health, 1);
+    }
 }

crates/obs/src/telemetry/local.rs

@@ -148,7 +148,7 @@ fn init_stdout_only(_config: &OtelConfig, logger_level: &str, is_production: boo
         .init();
 
     set_observability_metric_enabled(false);
-    counter!("rustfs.start.total").increment(1);
+    counter!("rustfs_start_total").increment(1);
     info!("Init stdout logging (level: {})", logger_level);
 
     OtelGuard {

crates/obs/src/telemetry/otel.rs

@@ -304,7 +304,7 @@ pub(super) fn init_observability_http(
         .with(metrics_layer)
         .init();
 
-    counter!("rustfs.start.total").increment(1);
+    counter!("rustfs_start_total").increment(1);
     info!(
         "Init observability (HTTP): trace='{}', metric='{}', log='{}'",
         trace_ep, metric_ep, log_ep

crates/s3select-api/src/query/execution.rs

@@ -52,7 +52,7 @@ impl Drop for PhaseTimer {
 
         if !std::thread::panicking() {
             metrics::histogram!(
-                "rustfs.s3select.phase.duration.seconds",
+                "rustfs_s3select_phase_duration_seconds",
                 "phase" => self.phase_name
             )
             .record(duration.as_secs_f64());
@@ -171,7 +171,7 @@ impl QueryStateMachine {
     fn record_phase_timestamp(&self, phase: &'static str, event: &'static str) {
         let elapsed = self.start.elapsed();
         metrics::histogram!(
-            "rustfs.s3select.phase.timestamp.seconds",
+            "rustfs_s3select_phase_timestamp_seconds",
             "phase" => phase,
             "event" => event
         )