feat: update gateway pool and usage flows

This commit is contained in:
fawney19
2026-05-12 21:05:11 +08:00
parent 38012c62ff
commit d1a47c068e
18 changed files with 582 additions and 328 deletions

View File

@@ -32,7 +32,8 @@ use crate::handlers::shared::provider_pool::{
admin_provider_pool_cache_affinity_enabled, admin_provider_pool_config_from_config_value,
};
use crate::handlers::shared::provider_pool::{
try_claim_admin_provider_pool_key, AdminProviderPoolConfig, AdminProviderPoolRuntimeState,
read_admin_provider_pool_key_cooldown_reason, AdminProviderPoolConfig,
AdminProviderPoolRuntimeState,
};
use crate::handlers::shared::{
parse_catalog_auth_config_json, provider_key_health_summary,
@@ -537,7 +538,7 @@ impl<'a> PoolKeyCursor<'a> {
async fn next_queued_candidate(&mut self) -> Option<EligibleLocalExecutionCandidate> {
while let Some(candidate) = self.queued_candidates.pop_front() {
let mut candidate = candidate;
if !self.attach_pool_key_lease(&mut candidate).await {
if self.skip_candidate_if_runtime_cooldown(&candidate).await {
continue;
}
candidate.orchestration.pool_key_index = Some(self.next_pool_key_index);
@@ -548,43 +549,42 @@ impl<'a> PoolKeyCursor<'a> {
None
}
async fn attach_pool_key_lease(
async fn skip_candidate_if_runtime_cooldown(
&mut self,
candidate: &mut EligibleLocalExecutionCandidate,
candidate: &EligibleLocalExecutionCandidate,
) -> bool {
if !pool_key_order_requires_exclusive_lease(&self.pool_key_order) {
return true;
}
let owner = self.state.app().tunnel.local_instance_id();
match try_claim_admin_provider_pool_key(
match read_admin_provider_pool_key_cooldown_reason(
self.state.app().runtime_state.as_ref(),
candidate.candidate.provider_id.as_str(),
candidate.candidate.key_id.as_str(),
owner,
)
.await
{
Ok(Some(lease)) => {
candidate.orchestration.pool_key_lease = Some(lease);
Ok(Some(_)) => {
self.record_skip_reason("pool_cooldown");
self.skipped_candidates
.push(SkippedLocalExecutionCandidate {
candidate: candidate.candidate.clone(),
skip_reason: "pool_cooldown",
transport: Some(candidate.transport.clone()),
ranking: candidate.ranking.clone(),
extra_data: None,
});
true
}
Ok(None) => {
self.record_skip_reason("pool_key_lease_busy");
false
}
Ok(None) => false,
Err(err) => {
warn!(
event_name = "pool_key_lease_claim_failed",
event_name = "pool_key_cooldown_check_failed",
log_type = "event",
provider_id = %candidate.candidate.provider_id,
endpoint_id = %candidate.candidate.endpoint_id,
model_id = %candidate.candidate.model_id,
key_id = %candidate.candidate.key_id,
error = ?err,
"gateway pool scheduler failed to claim pool key lease; scheduling key without lease"
"gateway pool scheduler failed to read pool key cooldown; scheduling key"
);
true
false
}
}
}
@@ -707,13 +707,6 @@ impl<'a> PoolKeyCursor<'a> {
}
}
fn pool_key_order_requires_exclusive_lease(order: &StoredPoolKeyCandidateOrder) -> bool {
!matches!(
order,
StoredPoolKeyCandidateOrder::CacheAffinity | StoredPoolKeyCandidateOrder::SingleAccount
)
}
fn pool_candidate_transport_policy_facts(
candidate: &aether_scheduler_core::SchedulerMinimalCandidateSelectionCandidate,
) -> CandidateTransportPolicyFacts<'_> {
@@ -1137,8 +1130,7 @@ fn apply_ai_pool_orchestration(
mod tests {
use super::{
apply_local_execution_pool_scheduler_with_runtime_map, build_pool_catalog_key_context,
pool_config_for_candidate, pool_key_order_requires_exclusive_lease, PoolCatalogKeyContext,
PoolKeyCursor,
pool_config_for_candidate, PoolCatalogKeyContext, PoolKeyCursor,
};
use crate::ai_serving::{
apply_local_runtime_candidate_terminal_reason, EligibleLocalExecutionCandidate,
@@ -1146,8 +1138,7 @@ mod tests {
};
use crate::data::GatewayDataState;
use crate::handlers::shared::provider_pool::{
record_admin_provider_pool_error, release_admin_provider_pool_key_lease,
try_claim_admin_provider_pool_key, AdminProviderPoolRuntimeState,
record_admin_provider_pool_error, AdminProviderPoolRuntimeState,
};
use crate::orchestration::LocalExecutionCandidateMetadata;
use crate::{AppState, LocalExecutionRuntimeMissDiagnostic};
@@ -1901,24 +1892,6 @@ mod tests {
assert_eq!(lru_cursor.pool_key_order, StoredPoolKeyCandidateOrder::Lru);
}
#[test]
fn pool_key_lease_exclusivity_follows_distribution_mode() {
assert!(!pool_key_order_requires_exclusive_lease(
&StoredPoolKeyCandidateOrder::CacheAffinity
));
assert!(!pool_key_order_requires_exclusive_lease(
&StoredPoolKeyCandidateOrder::SingleAccount
));
assert!(pool_key_order_requires_exclusive_lease(
&StoredPoolKeyCandidateOrder::Lru
));
assert!(pool_key_order_requires_exclusive_lease(
&StoredPoolKeyCandidateOrder::LoadBalance {
seed: "seed".to_string()
}
));
}
#[test]
fn pool_key_cursor_records_runtime_miss_when_exhausted_without_returning_key() {
let app = AppState::new().expect("state should build");
@@ -1941,8 +1914,8 @@ mod tests {
);
let mut cursor = PoolKeyCursor::new(PlannerAppState::new(&app), group, None, None, None)
.with_runtime_miss_diagnostic(trace_id, true);
cursor.record_skip_reason("pool_key_lease_busy");
cursor.record_skip_reason("pool_key_lease_busy");
cursor.record_skip_reason("pool_cooldown");
cursor.record_skip_reason("pool_cooldown");
cursor.record_skip_reason("transport_snapshot_missing");
cursor.log_exhausted();
@@ -1953,11 +1926,11 @@ mod tests {
.expect("runtime miss diagnostic should exist");
assert_eq!(diagnostic.reason, "all_candidates_skipped");
assert_eq!(diagnostic.skipped_candidate_count, Some(1));
assert_eq!(diagnostic.skip_reasons.get("pool_key_lease_busy"), Some(&1));
assert_eq!(diagnostic.skip_reasons.get("pool_cooldown"), Some(&1));
}
#[tokio::test]
async fn pool_key_cursor_freezes_queued_candidates_after_window_refill() {
async fn pool_key_cursor_rechecks_cooldown_for_frozen_window_candidates() {
let app = AppState::new().expect("state should build");
let provider_config = Some(json!({
"pool_advanced": {
@@ -2009,33 +1982,24 @@ mod tests {
let second = cursor
.next_key()
.await
.expect("second key should keep the frozen queue order");
assert_eq!(second.candidate.key_id, "key-b");
.expect("second key should skip the cooled-down frozen key");
assert_eq!(second.candidate.key_id, "key-c");
assert_eq!(second.orchestration.pool_key_index, Some(1));
let third = cursor
.next_key()
.await
.expect("third key should keep the frozen queue order");
assert_eq!(third.candidate.key_id, "key-c");
assert_eq!(third.orchestration.pool_key_index, Some(2));
let skipped = cursor.take_skipped_candidates();
assert!(skipped.is_empty());
assert_eq!(
skipped
.iter()
.map(|item| (item.candidate.key_id.as_str(), item.skip_reason))
.collect::<Vec<_>>(),
vec![("key-b", "pool_cooldown")]
);
}
#[tokio::test]
async fn pool_key_cursor_skips_busy_lease_and_claims_returned_key() {
async fn pool_key_cursor_allows_parallel_requests_to_use_same_healthy_key() {
let app = AppState::new().expect("state should build");
let provider_config = Some(json!({ "pool_advanced": { "lru_enabled": true } }));
let held_lease = try_claim_admin_provider_pool_key(
app.runtime_state.as_ref(),
"provider-pool",
"key-a",
"test-owner",
)
.await
.expect("lease claim should not fail")
.expect("first key should be claimed by test");
let group = sample_eligible_candidate(
"provider-pool",
"endpoint-1",
@@ -2052,87 +2016,19 @@ mod tests {
10,
provider_config.clone(),
),
sample_eligible_candidate("provider-pool", "endpoint-1", "key-b", 10, provider_config),
]);
let candidate = cursor
.next_key()
.await
.expect("cursor should skip busy key and return next key");
assert_eq!(candidate.candidate.key_id, "key-b");
assert_eq!(candidate.orchestration.pool_key_index, Some(0));
assert_eq!(
cursor.skip_reason_counts.get("pool_key_lease_busy"),
Some(&1)
);
let returned_lease = candidate
.orchestration
.pool_key_lease
.clone()
.expect("returned key should carry its lease");
assert!(
try_claim_admin_provider_pool_key(
app.runtime_state.as_ref(),
sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"key-b",
"other-owner",
)
.await
.expect("second claim should not fail")
.is_none(),
"returned key should remain leased until execution releases it"
);
release_admin_provider_pool_key_lease(app.runtime_state.as_ref(), &held_lease)
.await
.expect("held lease release should not fail");
release_admin_provider_pool_key_lease(app.runtime_state.as_ref(), &returned_lease)
.await
.expect("returned lease release should not fail");
}
#[tokio::test]
async fn pool_key_cursor_allows_busy_lease_for_cache_affinity_mode() {
let app = AppState::new().expect("state should build");
let provider_config = Some(json!({
"pool_advanced": {
"scheduling_presets": [
{"preset": "cache_affinity", "enabled": true}
]
}
}));
let held_lease = try_claim_admin_provider_pool_key(
app.runtime_state.as_ref(),
"provider-pool",
"key-a",
"test-owner",
)
.await
.expect("lease claim should not fail")
.expect("first key should be claimed by test");
let group = sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"pool-group",
10,
provider_config.clone(),
);
let mut cursor = PoolKeyCursor::new(PlannerAppState::new(&app), group, None, None, None);
cursor.queued_candidates = VecDeque::from([
sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"key-a",
10,
provider_config.clone(),
),
sample_eligible_candidate("provider-pool", "endpoint-1", "key-b", 10, provider_config),
]);
let candidate = cursor
.next_key()
.await
.expect("cache-affinity cursor should allow repeated busy key");
.expect("cursor should return the first healthy key");
assert_eq!(candidate.candidate.key_id, "key-a");
assert_eq!(candidate.orchestration.pool_key_index, Some(0));
assert!(candidate.orchestration.pool_key_lease.is_none());
@@ -2140,9 +2036,82 @@ mod tests {
.skip_reason_counts
.contains_key("pool_key_lease_busy"));
release_admin_provider_pool_key_lease(app.runtime_state.as_ref(), &held_lease)
let group = sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"pool-group",
10,
provider_config,
);
let mut second_cursor =
PoolKeyCursor::new(PlannerAppState::new(&app), group, None, None, None);
second_cursor.queued_candidates = VecDeque::from([
sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"key-a",
10,
Some(json!({ "pool_advanced": { "lru_enabled": true } })),
),
sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"key-b",
10,
Some(json!({ "pool_advanced": { "lru_enabled": true } })),
),
]);
let second_candidate = second_cursor
.next_key()
.await
.expect("held lease release should not fail");
.expect("second request should also be allowed to pick the same healthy key");
assert_eq!(second_candidate.candidate.key_id, "key-a");
assert!(second_candidate.orchestration.pool_key_lease.is_none());
}
#[tokio::test]
async fn pool_key_cursor_skips_key_after_account_cooldown_is_recorded() {
let app = AppState::new().expect("state should build");
let provider_config = Some(json!({ "pool_advanced": { "lru_enabled": true } }));
let group = sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"pool-group",
10,
provider_config.clone(),
);
let pool_config = pool_config_for_candidate(&group).expect("pool config should parse");
let mut cursor = PoolKeyCursor::new(PlannerAppState::new(&app), group, None, None, None);
cursor.queued_candidates = VecDeque::from([
sample_eligible_candidate(
"provider-pool",
"endpoint-1",
"key-a",
10,
provider_config.clone(),
),
sample_eligible_candidate("provider-pool", "endpoint-1", "key-b", 10, provider_config),
]);
record_admin_provider_pool_error(
app.runtime_state.as_ref(),
"provider-pool",
"key-a",
&pool_config,
429,
None,
None,
)
.await;
let candidate = cursor
.next_key()
.await
.expect("cursor should skip cooled-down key and return next key");
assert_eq!(candidate.candidate.key_id, "key-b");
assert_eq!(candidate.orchestration.pool_key_index, Some(0));
assert!(candidate.orchestration.pool_key_lease.is_none());
assert_eq!(cursor.skip_reason_counts.get("pool_cooldown"), Some(&1));
}
#[tokio::test]
@@ -2177,21 +2146,7 @@ mod tests {
provider_config.clone(),
);
let pool_config = pool_config_for_candidate(&group).expect("pool config should parse");
let mut held_leases = Vec::new();
for key_id in ["key-00000", "key-00001"] {
held_leases.push(
try_claim_admin_provider_pool_key(
app.runtime_state.as_ref(),
"provider-pool",
key_id,
"large-pool-test",
)
.await
.expect("lease claim should not fail")
.expect("test key should claim"),
);
}
for key_id in ["key-00002", "key-00003"] {
record_admin_provider_pool_error(
app.runtime_state.as_ref(),
"provider-pool",
@@ -2219,22 +2174,16 @@ mod tests {
);
let mut returned_ids = Vec::new();
let mut returned_leases = Vec::new();
for _ in 0..10 {
let candidate = cursor
.next_key()
.await
.expect("large pool should return first page candidates");
returned_ids.push(candidate.candidate.key_id.clone());
returned_leases.push(
candidate
.orchestration
.pool_key_lease
.expect("lru pool candidate should carry exclusive lease"),
);
assert!(candidate.orchestration.pool_key_lease.is_none());
}
assert_eq!(returned_ids.first().map(String::as_str), Some("key-00004"));
assert_eq!(returned_ids.last().map(String::as_str), Some("key-00013"));
assert_eq!(returned_ids.first().map(String::as_str), Some("key-00002"));
assert_eq!(returned_ids.last().map(String::as_str), Some("key-00011"));
assert_eq!(cursor.scanned_keys, 64);
assert!(
cursor.queued_candidates.len() <= cursor.window_size as usize,
@@ -2251,39 +2200,18 @@ mod tests {
None,
)
.await;
held_leases.push(
try_claim_admin_provider_pool_key(
app.runtime_state.as_ref(),
"provider-pool",
"key-00015",
"large-pool-test",
)
.await
.expect("dynamic lease claim should not fail")
.expect("dynamic busy key should claim"),
);
let candidate = cursor
.next_key()
.await
.expect("cursor should keep the frozen window despite later runtime changes");
assert_eq!(candidate.candidate.key_id, "key-00014");
assert_eq!(candidate.candidate.key_id, "key-00012");
returned_ids.push(candidate.candidate.key_id.clone());
returned_leases.push(
candidate
.orchestration
.pool_key_lease
.expect("lru pool candidate should carry exclusive lease"),
);
assert!(candidate.orchestration.pool_key_lease.is_none());
while let Some(candidate) = cursor.next_key().await {
returned_ids.push(candidate.candidate.key_id.clone());
returned_leases.push(
candidate
.orchestration
.pool_key_lease
.expect("lru pool candidate should carry exclusive lease"),
);
assert!(candidate.orchestration.pool_key_lease.is_none());
}
let max_returned_windows = aether_dispatch_core::DEFAULT_POOL_MAX_SCAN
@@ -2293,34 +2221,23 @@ mod tests {
<= (max_returned_windows * aether_dispatch_core::DEFAULT_POOL_WINDOW_SIZE) as usize,
"cursor should only return bounded frozen windows per request"
);
assert_eq!(returned_ids.len(), 125);
assert_eq!(returned_ids.len(), 127);
assert_eq!(returned_ids.last().map(String::as_str), Some("key-00463"));
assert_eq!(cursor.scanned_keys, 512);
assert_eq!(
cursor.skip_reason_counts.get("pool_key_lease_busy"),
Some(&3)
);
assert_eq!(cursor.skip_reason_counts.get("pool_cooldown"), Some(&2));
for skipped in ["key-00000", "key-00001", "key-00002", "key-00003"] {
assert_eq!(cursor.skip_reason_counts.get("pool_cooldown"), Some(&3));
assert!(!cursor
.skip_reason_counts
.contains_key("pool_key_lease_busy"));
for skipped in ["key-00000", "key-00001", "key-00014"] {
assert!(
!returned_ids.iter().any(|key_id| key_id == skipped),
"{skipped} should have been skipped"
);
}
assert!(
returned_ids.iter().any(|key_id| key_id == "key-00014"),
"key-00014 was already frozen in the current window"
returned_ids.iter().any(|key_id| key_id == "key-00015"),
"key-00015 should not be blocked by request-scoped leases"
);
assert!(
!returned_ids.iter().any(|key_id| key_id == "key-00015"),
"key-00015 should still be skipped because exclusive lease is checked at dispatch time"
);
for lease in held_leases.into_iter().chain(returned_leases) {
release_admin_provider_pool_key_lease(app.runtime_state.as_ref(), &lease)
.await
.expect("lease release should not fail");
}
}
#[test]