pub struct SchedulerLoadSnapshot {Show 19 fields
pub dp_rank: i32,
pub num_running_reqs: i32,
pub num_waiting_reqs: i32,
pub num_waiting_uncached_tokens: i32,
pub num_total_reqs: i32,
pub num_used_tokens: i32,
pub max_total_num_tokens: i32,
pub token_usage: f64,
pub gen_throughput: f64,
pub cache_hit_rate: f64,
pub utilization: f64,
pub max_running_requests: i32,
pub memory: Option<EngineMemoryMetricsSnapshot>,
pub queues: Option<EngineQueueMetricsSnapshot>,
pub kv_transfer_latency_ms: Option<f64>,
pub kv_transfer_speed_gb_s: Option<f64>,
pub prefill_queue_reqs: Option<i32>,
pub decode_queue_reqs: Option<i32>,
pub disagg_mode: Option<String>,
}Expand description
Per-DP-rank load snapshot from a backend.
Contains core metrics from the sglang /v1/loads endpoint or GetLoads gRPC RPC.
Each snapshot represents one data-parallel rank’s scheduler state.
Fields§
§dp_rank: i32§num_running_reqs: i32§num_waiting_reqs: i32§num_waiting_uncached_tokens: i32Queued token-work: waiting-queue tokens not yet served from cache. 0 when the backend does not report it — callers degrade gracefully.
num_total_reqs: i32§num_used_tokens: i32§max_total_num_tokens: i32§token_usage: f64Token usage ratio (0.0–1.0).
gen_throughput: f64§cache_hit_rate: f64§utilization: f64§max_running_requests: i32§memory: Option<EngineMemoryMetricsSnapshot>§queues: Option<EngineQueueMetricsSnapshot>§kv_transfer_latency_ms: Option<f64>PD disaggregation signals, populated only when the backend reports a
disagg section. None for HTTP or older engines. Canonical schema
other engines map into; SGLang derives the queue depths from its
per-stage DisaggregationMetrics counters.
kv_transfer_speed_gb_s: Option<f64>§prefill_queue_reqs: Option<i32>§decode_queue_reqs: Option<i32>§disagg_mode: Option<String>“prefill”, “decode”, or “null” as reported by the backend.
Trait Implementations§
Source§impl Clone for SchedulerLoadSnapshot
impl Clone for SchedulerLoadSnapshot
Source§fn clone(&self) -> SchedulerLoadSnapshot
fn clone(&self) -> SchedulerLoadSnapshot
Returns a duplicate of the value. Read more
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
Performs copy-assignment from
source. Read moreSource§impl Debug for SchedulerLoadSnapshot
impl Debug for SchedulerLoadSnapshot
Source§impl Default for SchedulerLoadSnapshot
impl Default for SchedulerLoadSnapshot
Source§fn default() -> SchedulerLoadSnapshot
fn default() -> SchedulerLoadSnapshot
Returns the “default value” for a type. Read more
Source§impl<'de> Deserialize<'de> for SchedulerLoadSnapshotwhere
SchedulerLoadSnapshot: Default,
impl<'de> Deserialize<'de> for SchedulerLoadSnapshotwhere
SchedulerLoadSnapshot: Default,
Source§fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
__D: Deserializer<'de>,
fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
__D: Deserializer<'de>,
Deserialize this value from the given Serde deserializer. Read more
Auto Trait Implementations§
impl Freeze for SchedulerLoadSnapshot
impl RefUnwindSafe for SchedulerLoadSnapshot
impl Send for SchedulerLoadSnapshot
impl Sync for SchedulerLoadSnapshot
impl Unpin for SchedulerLoadSnapshot
impl UnsafeUnpin for SchedulerLoadSnapshot
impl UnwindSafe for SchedulerLoadSnapshot
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more