BCAI/easyai-ai-gateway
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat(worker): 实现集群限流与自适应负载

Browse Source 
保留平台模型 RPM、TPM 和并发策略语义，增加 PostgreSQL 集群级租约、饱和候选重选和多平台自动负载，避免突发任务固定等待首个平台。\n\n新增 Worker 实时负载采样、自适应 active/heavy 容量、心跳与管理端指标，并扩展本地 acceptance runner，覆盖三 Worker、同模型三平台 2/4/6 并发和 48 个带图视频突发任务。\n\n验证：go test ./...、go vet ./...、PostgreSQL 跨 Store 集成测试、gofmt、bash -n、ShellCheck 及本地集群 provider-burst 验收通过；48/48 成功，无越限、重复提交、重复计费、重复回调或终态资源泄漏。
This commit is contained in:
wangbo
2026-08-03 00:13:46 +08:00
parent 9a01fd4657
commit c28bf74230
52 changed files with 3700 additions and 272 deletions
Download Patch File Download Diff File Expand all files Collapse all files
apps/api/internal/runner/admission_test.go
+33 -11
View File
@@ -57,7 +57,7 @@ func TestDistributedAdmissionModelTypeBoundary(t *testing.T) {
}
}
func TestAcceptanceAdmissionScopesUseWorkerCapacityInsteadOfSupplierConcurrency(t *testing.T) {
func TestAcceptanceAdmissionScopesIsolateRunWithoutChangingLimits(t *testing.T) {
input := []store.AdmissionScope{{
ScopeType: "platform_model",
ScopeKey: "model-1",
@@ -70,15 +70,15 @@ func TestAcceptanceAdmissionScopesUseWorkerCapacityInsteadOfSupplierConcurrency(
ConcurrentLimit: 48,
}}
got := acceptanceAdmissionScopes(store.GatewayTask{RunMode: "acceptance"}, input)
got := acceptanceAdmissionScopes(store.GatewayTask{RunMode: "acceptance", AcceptanceRunID: "run-1"}, input)
if got[0].ConcurrentLimit != 0 {
t.Fatalf("protocol-emulated acceptance must defer to worker capacity, got %+v", got[0])
if got[0].ConcurrentLimit != 10 || got[0].ScopeKey != "acceptance:run-1:model-1" {
t.Fatalf("protocol-emulated acceptance must preserve the limit in an isolated scope, got %+v", got[0])
}
if got[0].QueueLimit != acceptanceQueueLimit || got[0].MaxWaitSeconds != acceptanceQueueMaxWait {
t.Fatalf("acceptance must enable a bounded queue, got %+v", got[0])
}
if got[1].ConcurrentLimit != 48 {
if got[1].ConcurrentLimit != 48 || got[1].ScopeKey != "acceptance:run-1:global" {
t.Fatalf("acceptance worker capacity changed, got %+v", got[1])
}
if input[0].QueueLimit != 0 || input[0].MaxWaitSeconds != 0 {
@@ -100,15 +100,15 @@ func TestAcceptanceCanaryPreservesProductionConcurrency(t *testing.T) {
}
}
func TestAcceptanceInfrastructureReservationsOnlyRemoveSupplierConcurrency(t *testing.T) {
func TestAcceptanceInfrastructureReservationsIsolateEveryMetric(t *testing.T) {
input := []store.RateLimitReservation{
{ScopeType: "platform_model", Metric: "concurrent", Limit: 10},
{ScopeType: "platform_model", Metric: "rpm", Limit: 600},
{ScopeType: "user_group", Metric: "concurrent", Limit: 20},
{ScopeType: "platform_model", ScopeKey: "model-1", Metric: "concurrent", Limit: 10},
{ScopeType: "platform_model", ScopeKey: "model-1", Metric: "rpm", Limit: 600},
{ScopeType: "user_group", ScopeKey: "group-1", Metric: "concurrent", Limit: 20},
}
got := acceptanceInfrastructureReservations(store.GatewayTask{RunMode: "acceptance"}, input)
if len(got) != 2 || got[0].Metric != "rpm" || got[1].ScopeType != "user_group" {
got := acceptanceInfrastructureReservations(store.GatewayTask{RunMode: "acceptance", AcceptanceRunID: "run-1"}, input)
if len(got) != 3 || got[0].ScopeKey != "acceptance:run-1:model-1" || got[1].ScopeKey != "acceptance:run-1:model-1" || got[2].ScopeKey != "acceptance:run-1:group-1" {
t.Fatalf("unexpected acceptance reservations: %+v", got)
}
canary := acceptanceInfrastructureReservations(store.GatewayTask{RunMode: "acceptance_canary"}, input)
@@ -192,6 +192,28 @@ func TestPinCandidatesToTaskAdmissionPreservesWaitingCandidate(t *testing.T) {
}
}
func TestPinCandidatesToTaskAdmissionAllowsRequestedReselection(t *testing.T) {
input := []store.RuntimeModelCandidate{
{PlatformID: "platform-a", PlatformModelID: "model-a"},
{PlatformID: "platform-b", PlatformModelID: "model-b"},
}
admission := &store.TaskAdmission{
Status: "waiting",
PlatformID: "platform-b",
PlatformModelID: "model-b",
ReselectRequestedAt: time.Now(),
}
got, pinned := pinCandidatesToTaskAdmission(input, admission)
if pinned {
t.Fatalf("reselection request unexpectedly pinned the old candidate: %+v", got)
}
if got[0].PlatformModelID != "model-a" {
t.Fatalf("reselection changed the sorted candidate order: %+v", got)
}
}
func TestPinCandidatesToTaskAdmissionIgnoresMissingCandidate(t *testing.T) {
input := []store.RuntimeModelCandidate{
{PlatformID: "platform-a", PlatformModelID: "model-a"},