Fix provider pool exhaustion scheduling

This commit is contained in:
fawney19
2026-05-20 20:13:07 +08:00
parent d0981c2fd5
commit c972bbd397
9 changed files with 424 additions and 48 deletions

View File

@@ -380,6 +380,81 @@ mod tests {
));
}
#[test]
fn provider_quota_exhaustion_snapshot_expires_after_reset_at() {
let now = std::time::SystemTime::now()
.duration_since(std::time::UNIX_EPOCH)
.expect("system time should be after unix epoch")
.as_secs();
let mut expired = sample_key(None);
expired.status_snapshot = Some(json!({
"quota": {
"version": 2,
"provider_type": "codex",
"code": "exhausted",
"exhausted": true,
"updated_at": now.saturating_sub(600),
"windows": [{
"code": "5h",
"used_ratio": 1.0,
"reset_at": now.saturating_sub(60),
"is_exhausted": true
}]
}
}));
assert!(!provider_pool_key_account_quota_exhausted(
&expired, "codex"
));
let mut active = sample_key(None);
active.status_snapshot = Some(json!({
"quota": {
"version": 2,
"provider_type": "codex",
"code": "exhausted",
"exhausted": true,
"updated_at": now,
"windows": [{
"code": "5h",
"used_ratio": 1.0,
"reset_at": now.saturating_add(3600),
"is_exhausted": true
}]
}
}));
assert!(provider_pool_key_account_quota_exhausted(&active, "codex"));
}
#[test]
fn provider_quota_exhaustion_metadata_expires_after_reset_at() {
let now = std::time::SystemTime::now()
.duration_since(std::time::UNIX_EPOCH)
.expect("system time should be after unix epoch")
.as_secs();
assert!(!provider_pool_key_account_quota_exhausted(
&sample_key(Some(json!({
"codex": {
"updated_at": now.saturating_sub(600),
"primary_used_percent": 100.0,
"primary_reset_at": now.saturating_sub(60)
}
}))),
"codex",
));
assert!(provider_pool_key_account_quota_exhausted(
&sample_key(Some(json!({
"codex": {
"updated_at": now,
"primary_used_percent": 100.0,
"primary_reset_at": now.saturating_add(3600)
}
}))),
"codex",
));
}
#[test]
fn grok_quota_tier_boundaries_match_pool_modes() {
assert_eq!(

View File

@@ -10,8 +10,9 @@ use crate::provider::{
ProviderPoolMemberInput,
};
use crate::quota::{
provider_pool_json_bool, provider_pool_json_f64, provider_pool_metadata_bucket,
provider_pool_quota_snapshot_exhausted_decision,
provider_pool_current_unix_secs, provider_pool_json_bool, provider_pool_json_f64,
provider_pool_metadata_bucket, provider_pool_quota_snapshot_exhausted_decision,
provider_pool_reset_deadline_elapsed, provider_pool_timestamp_unix_secs,
};
use crate::quota_refresh::ProviderPoolQuotaRequestSpec;
@@ -255,6 +256,19 @@ fn infer_chatgpt_web_image_quota_limit(
}
pub(crate) fn quota_exhausted_from_bucket(bucket: &Map<String, Value>) -> bool {
if provider_pool_current_unix_secs().is_some_and(|now| {
let mut image_quota = Map::new();
if let Some(value) = bucket.get("image_quota_reset_at") {
image_quota.insert("reset_at".to_string(), value.clone());
}
provider_pool_reset_deadline_elapsed(
&image_quota,
provider_pool_timestamp_unix_secs(bucket.get("updated_at")),
now,
)
}) {
return false;
}
if provider_pool_json_bool(bucket.get("image_quota_blocked")) == Some(true) {
return true;
}

View File

@@ -10,8 +10,9 @@ use crate::provider::{
ProviderPoolMemberInput,
};
use crate::quota::{
provider_pool_json_bool, provider_pool_json_f64, provider_pool_metadata_bucket,
provider_pool_quota_snapshot_exhausted_decision,
provider_pool_current_unix_secs, provider_pool_json_bool, provider_pool_json_f64,
provider_pool_metadata_bucket, provider_pool_quota_snapshot_exhausted_decision,
provider_pool_reset_deadline_elapsed, provider_pool_timestamp_unix_secs,
};
use crate::quota_refresh::ProviderPoolQuotaRequestSpec;
@@ -121,6 +122,37 @@ pub fn build_codex_pool_quota_request(
})
}
fn codex_window_reset_elapsed(bucket: &Map<String, Value>, prefix: &str) -> bool {
let Some(now_unix_secs) = provider_pool_current_unix_secs() else {
return false;
};
let mut window = Map::new();
for (target, source) in [
("reset_at", format!("{prefix}_reset_at")),
("next_reset_at", format!("{prefix}_next_reset_at")),
("reset_seconds", format!("{prefix}_reset_seconds")),
(
"reset_after_seconds",
format!("{prefix}_reset_after_seconds"),
),
] {
if let Some(value) = bucket.get(source.as_str()) {
window.insert(target.to_string(), value.clone());
}
}
provider_pool_reset_deadline_elapsed(
&window,
provider_pool_timestamp_unix_secs(bucket.get("updated_at")),
now_unix_secs,
)
}
fn codex_window_used_percent_exhausted(bucket: &Map<String, Value>, prefix: &str) -> bool {
let used_percent_key = format!("{prefix}_used_percent");
provider_pool_json_f64(bucket.get(used_percent_key.as_str()))
.is_some_and(|value| value >= 100.0 && !codex_window_reset_elapsed(bucket, prefix))
}
pub(crate) fn quota_exhausted_from_bucket(bucket: &Map<String, Value>) -> bool {
if provider_pool_json_bool(bucket.get("credits_unlimited")) == Some(true) {
return false;
@@ -130,7 +162,6 @@ pub(crate) fn quota_exhausted_from_bucket(bucket: &Map<String, Value>) -> bool {
if !has_window_data && provider_pool_json_bool(bucket.get("has_credits")) == Some(false) {
return true;
}
provider_pool_json_f64(bucket.get("primary_used_percent")).is_some_and(|value| value >= 100.0)
|| provider_pool_json_f64(bucket.get("secondary_used_percent"))
.is_some_and(|value| value >= 100.0)
codex_window_used_percent_exhausted(bucket, "primary")
|| codex_window_used_percent_exhausted(bucket, "secondary")
}

View File

@@ -7,8 +7,9 @@ use crate::provider::{
ProviderPoolMemberInput,
};
use crate::quota::{
provider_pool_json_bool, provider_pool_json_f64, provider_pool_metadata_bucket,
provider_pool_quota_snapshot_exhausted_decision,
provider_pool_current_unix_secs, provider_pool_json_bool, provider_pool_json_f64,
provider_pool_metadata_bucket, provider_pool_quota_snapshot_exhausted_decision,
provider_pool_reset_deadline_elapsed, provider_pool_timestamp_unix_secs,
};
pub const GROK_QUOTA_WINDOWS_BASIC: &[(&str, &str)] = &[("quota_fast", "fast")];
@@ -191,6 +192,8 @@ pub(crate) fn quota_exhausted_from_bucket(bucket: &Map<String, Value>) -> bool {
let mut model_count = 0usize;
let mut exhausted_count = 0usize;
let now_unix_secs = provider_pool_current_unix_secs();
let bucket_updated_at = provider_pool_timestamp_unix_secs(bucket.get("updated_at"));
for (model_key, item) in models.iter() {
if !supported_mode_keys.is_empty() && !supported_mode_keys.contains(&model_key.as_str()) {
continue;
@@ -207,12 +210,14 @@ pub(crate) fn quota_exhausted_from_bucket(bucket: &Map<String, Value>) -> bool {
continue;
}
model_count += 1;
if provider_pool_json_bool(item.get("is_exhausted")) == Some(true)
let quota_exhausted = provider_pool_json_bool(item.get("is_exhausted")) == Some(true)
|| provider_pool_json_f64(item.get("used_percent")).is_some_and(|value| value >= 100.0)
|| provider_pool_json_f64(item.get("remaining")).is_some_and(|value| value <= 0.0)
|| provider_pool_json_f64(item.get("remaining_fraction"))
.is_some_and(|value| value <= 0.0)
{
.is_some_and(|value| value <= 0.0);
let reset_elapsed = now_unix_secs
.is_some_and(|now| provider_pool_reset_deadline_elapsed(item, bucket_updated_at, now));
if quota_exhausted && !reset_elapsed {
exhausted_count += 1;
}
}

View File

@@ -11,8 +11,9 @@ use crate::provider::{
ProviderPoolMemberInput,
};
use crate::quota::{
provider_pool_json_f64, provider_pool_metadata_bucket,
provider_pool_quota_snapshot_exhausted_decision,
provider_pool_current_unix_secs, provider_pool_json_f64, provider_pool_metadata_bucket,
provider_pool_quota_snapshot_exhausted_decision, provider_pool_reset_deadline_elapsed,
provider_pool_timestamp_unix_secs,
};
use crate::quota_refresh::ProviderPoolQuotaRequestSpec;
@@ -151,6 +152,15 @@ fn normalize_kiro_version(value: &str) -> &str {
}
pub(crate) fn quota_exhausted_from_bucket(bucket: &Map<String, Value>) -> bool {
if provider_pool_current_unix_secs().is_some_and(|now| {
provider_pool_reset_deadline_elapsed(
bucket,
provider_pool_timestamp_unix_secs(bucket.get("updated_at")),
now,
)
}) {
return false;
}
if provider_pool_json_f64(bucket.get("remaining")).is_some_and(|value| value <= 0.0) {
return true;
}

View File

@@ -1,3 +1,5 @@
use std::time::{SystemTime, UNIX_EPOCH};
use aether_data_contracts::repository::provider_catalog::StoredProviderCatalogKey;
use serde_json::{json, Map, Value};
@@ -135,7 +137,7 @@ pub(crate) fn provider_pool_json_f64(value: Option<&Value>) -> Option<f64> {
.filter(|value| value.is_finite())
}
fn provider_pool_timestamp_unix_secs(value: Option<&Value>) -> Option<u64> {
pub(crate) fn provider_pool_timestamp_unix_secs(value: Option<&Value>) -> Option<u64> {
let mut timestamp = provider_pool_json_f64(value)?;
if timestamp <= 0.0 {
return None;
@@ -146,6 +148,49 @@ fn provider_pool_timestamp_unix_secs(value: Option<&Value>) -> Option<u64> {
Some(timestamp as u64)
}
pub(crate) fn provider_pool_current_unix_secs() -> Option<u64> {
SystemTime::now()
.duration_since(UNIX_EPOCH)
.ok()
.map(|duration| duration.as_secs())
}
fn provider_pool_reset_deadline_unix_secs(
item: &Map<String, Value>,
fallback_observed_at: Option<u64>,
) -> Option<u64> {
provider_pool_timestamp_unix_secs(item.get("reset_at"))
.or_else(|| provider_pool_timestamp_unix_secs(item.get("next_reset_at")))
.or_else(|| {
let reset_seconds = provider_pool_json_f64(item.get("reset_seconds"))
.or_else(|| provider_pool_json_f64(item.get("reset_after_seconds")))?;
if reset_seconds < 0.0 {
return None;
}
let base = provider_pool_timestamp_unix_secs(item.get("observed_at"))
.or_else(|| provider_pool_timestamp_unix_secs(item.get("updated_at")))
.or(fallback_observed_at)?;
Some(base.saturating_add(reset_seconds.ceil() as u64))
})
}
pub(crate) fn provider_pool_reset_deadline_elapsed(
item: &Map<String, Value>,
fallback_observed_at: Option<u64>,
now_unix_secs: u64,
) -> bool {
provider_pool_reset_deadline_unix_secs(item, fallback_observed_at)
.is_some_and(|reset_at| reset_at <= now_unix_secs)
}
fn provider_pool_quota_window_is_exhausted(window: &Map<String, Value>) -> bool {
provider_pool_json_bool(window.get("is_exhausted"))
.or_else(|| {
provider_pool_json_f64(window.get("used_ratio")).map(|value| value >= 1.0 - 1e-6)
})
.unwrap_or(false)
}
fn provider_pool_quota_snapshot_matches_provider(
quota_snapshot: &Map<String, Value>,
provider_type: &str,
@@ -203,19 +248,45 @@ pub(crate) fn provider_pool_quota_snapshot_exhausted_decision(
}
let exhausted = provider_pool_json_bool(quota_snapshot.get("exhausted"))?;
if exhausted {
let windows_max_ratio = quota_snapshot
let now_unix_secs = provider_pool_current_unix_secs();
let snapshot_observed_at =
provider_pool_timestamp_unix_secs(quota_snapshot.get("observed_at"))
.or_else(|| provider_pool_timestamp_unix_secs(quota_snapshot.get("updated_at")));
if let Some(windows) = quota_snapshot
.get("windows")
.and_then(Value::as_array)
.filter(|w| !w.is_empty())
.and_then(|windows| {
windows
.iter()
.filter_map(Value::as_object)
.filter_map(|w| w.get("used_ratio"))
.filter_map(Value::as_f64)
.max_by(f64::total_cmp)
});
if windows_max_ratio.is_some_and(|ratio| ratio < 1.0 - 1e-6) {
.filter(|windows| !windows.is_empty())
{
let mut saw_exhausted_window = false;
let mut saw_active_exhausted_window = false;
let mut windows_max_ratio = None::<f64>;
for window in windows.iter().filter_map(Value::as_object) {
if let Some(ratio) = provider_pool_json_f64(window.get("used_ratio")) {
windows_max_ratio =
Some(windows_max_ratio.map_or(ratio, |current| current.max(ratio)));
}
if provider_pool_quota_window_is_exhausted(window) {
saw_exhausted_window = true;
let reset_elapsed = now_unix_secs.is_some_and(|now| {
provider_pool_reset_deadline_elapsed(window, snapshot_observed_at, now)
});
if !reset_elapsed {
saw_active_exhausted_window = true;
}
}
}
if saw_exhausted_window {
return Some(saw_active_exhausted_window);
}
if windows_max_ratio.is_some_and(|ratio| ratio < 1.0 - 1e-6) {
return Some(false);
}
} else if now_unix_secs.is_some_and(|now| {
provider_pool_reset_deadline_elapsed(quota_snapshot, snapshot_observed_at, now)
}) {
return Some(false);
}
}