fix: 停复机指令队列审查问题
All checks were successful
构建并部署到测试环境(无 SSH) / build-and-deploy (push) Successful in 9m24s

- 开关开启但分发器心跳缺失时回退为当场执行,避免指令无人消费
- 待确认指令先查上游状态再判断;执行期间被覆盖时把待确认带到新指令
- 同方向失败退避中重复写入不提前执行时间
- 领取容量把执行上限与本秒剩余 QPS 分开计算
- 队列复机共用单卡复机锁;执行上下文取消后按结果未知确认
- 消费者并行停止;读取最近调用时间失败时不覆盖;成功但终结日志失败按成功处理

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
邹进
2026-10-26 00:43:58 +09:00
parent 934a3a27a1
commit a4e3fe4d8f
9 changed files with 188 additions and 79 deletions

View File

@@ -42,29 +42,30 @@ func (d *Dispatcher) stopResumeLease() time.Duration {
// dispatchStopResume 在卡任务之前领取账户停复机指令:领取数 = min(停复机所在限流组本秒剩余, 执行上限 − 执行中)。
func (d *Dispatcher) dispatchStopResume(ctx context.Context, carrierID uint) {
params := d.ParamsFor(carrierID)
capacity := params.SRInflightLimit
limit := params.SRInflightLimit
processes := d.cfg.SubscriberProcesses
if processes <= 0 {
processes = 1
}
if consume := params.SRMaxInFlight * processes; consume > 0 && capacity > consume {
capacity = consume
if consume := params.SRMaxInFlight * processes; consume > 0 && limit > consume {
limit = consume
}
budget := -1
if d.budgets != nil {
ops := []string{constants.GatewayOpStop, constants.GatewayOpStart}
if groups, _, limited := d.budgets.Budget(ctx, carrierID, ops); limited {
for _, remaining := range groups {
if remaining < capacity {
capacity = remaining
if budget < 0 || remaining < budget {
budget = remaining
}
}
}
}
if capacity <= 0 {
if limit <= 0 || budget == 0 {
return
}
claimed, expired, err := d.store.ClaimStopResume(ctx, SRClaimInput{
CarrierID: carrierID, Lease: d.stopResumeLease(), Capacity: capacity, MinInterval: params.SRMinInterval,
CarrierID: carrierID, Lease: d.stopResumeLease(), InflightLimit: limit, Budget: budget, MinInterval: params.SRMinInterval,
TokenPrefix: d.instanceID + ":sr:" + strconv.FormatInt(time.Now().UnixNano(), 36),
})
if err != nil {