feat(worker): 实现集群限流与自适应负载

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:
2026-08-03 00:13:46 +08:00
parent 9a01fd4657
commit c28bf74230
52 changed files with 3700 additions and 272 deletions
@@ -6,19 +6,28 @@ import (
)
type AsyncWorkerCapacitySnapshot struct {
Capacity int
GlobalCapacity int
Desired int
HardLimit int
Capped bool
EnabledModels int
UnlimitedModels int
EnabledGroups int
UnlimitedGroups int
ModelDesired int
GroupDesired int
ActiveInstances int
InstanceID string
Capacity int
GlobalCapacity int
Desired int
HardLimit int
Capped bool
EnabledModels int
UnlimitedModels int
EnabledGroups int
UnlimitedGroups int
ModelDesired int
GroupDesired int
ActiveInstances int
InstanceID string
LoadMode string
LocalSafeCapacity int
LocalHeavyCapacity int
LocalActiveTasks int
LocalPreparingTasks int
LocalWaitingTasks int
LocalFinalizingTasks int
LocalPressureState string
LocalPressureReason string
}
func (s *Store) AsyncWorkerCapacity(ctx context.Context, hardLimit int) (AsyncWorkerCapacitySnapshot, error) {