feat(scheduler): 引入调度快照缓存与 outbox 回放

- 调度热路径优先读 Redis 快照,保留分组排序语义
- outbox 回放 + 全量重建纠偏,失败重试不推进水位
- 自动 Atlas 基线对齐并同步调度配置示例
This commit is contained in:
yangjianbo
2026-01-12 14:19:06 +08:00
parent b76cc583fb
commit 3141aa5144
20 changed files with 1785 additions and 83 deletions

View File

@@ -40,6 +40,7 @@ type GeminiMessagesCompatService struct {
accountRepo AccountRepository
groupRepo GroupRepository
cache GatewayCache
schedulerSnapshot *SchedulerSnapshotService
tokenProvider *GeminiTokenProvider
rateLimitService *RateLimitService
httpUpstream HTTPUpstream
@@ -51,6 +52,7 @@ func NewGeminiMessagesCompatService(
accountRepo AccountRepository,
groupRepo GroupRepository,
cache GatewayCache,
schedulerSnapshot *SchedulerSnapshotService,
tokenProvider *GeminiTokenProvider,
rateLimitService *RateLimitService,
httpUpstream HTTPUpstream,
@@ -61,6 +63,7 @@ func NewGeminiMessagesCompatService(
accountRepo: accountRepo,
groupRepo: groupRepo,
cache: cache,
schedulerSnapshot: schedulerSnapshot,
tokenProvider: tokenProvider,
rateLimitService: rateLimitService,
httpUpstream: httpUpstream,
@@ -105,12 +108,6 @@ func (s *GeminiMessagesCompatService) SelectAccountForModelWithExclusions(ctx co
// gemini 分组支持混合调度(包含启用了 mixed_scheduling 的 antigravity 账户)
// 注意:强制平台模式不走混合调度
useMixedScheduling := platform == PlatformGemini && !hasForcePlatform
var queryPlatforms []string
if useMixedScheduling {
queryPlatforms = []string{PlatformGemini, PlatformAntigravity}
} else {
queryPlatforms = []string{platform}
}
cacheKey := "gemini:" + sessionHash
@@ -118,7 +115,7 @@ func (s *GeminiMessagesCompatService) SelectAccountForModelWithExclusions(ctx co
accountID, err := s.cache.GetSessionAccountID(ctx, derefGroupID(groupID), cacheKey)
if err == nil && accountID > 0 {
if _, excluded := excludedIDs[accountID]; !excluded {
account, err := s.accountRepo.GetByID(ctx, accountID)
account, err := s.getSchedulableAccount(ctx, accountID)
// 检查账号是否有效原生平台直接匹配antigravity 需要启用混合调度
if err == nil && account.IsSchedulableForModel(requestedModel) && (requestedModel == "" || s.isModelSupportedByAccount(account, requestedModel)) {
valid := false
@@ -149,22 +146,16 @@ func (s *GeminiMessagesCompatService) SelectAccountForModelWithExclusions(ctx co
}
// 查询可调度账户(强制平台模式:优先按分组查找,找不到再查全部)
var accounts []Account
var err error
if groupID != nil {
accounts, err = s.accountRepo.ListSchedulableByGroupIDAndPlatforms(ctx, *groupID, queryPlatforms)
accounts, err := s.listSchedulableAccountsOnce(ctx, groupID, platform, hasForcePlatform)
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}
// 强制平台模式下,分组中找不到账户时回退查询全部
if len(accounts) == 0 && groupID != nil && hasForcePlatform {
accounts, err = s.listSchedulableAccountsOnce(ctx, nil, platform, hasForcePlatform)
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}
// 强制平台模式下,分组中找不到账户时回退查询全部
if len(accounts) == 0 && hasForcePlatform {
accounts, err = s.accountRepo.ListSchedulableByPlatforms(ctx, queryPlatforms)
}
} else {
accounts, err = s.accountRepo.ListSchedulableByPlatforms(ctx, queryPlatforms)
}
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}
var selected *Account
@@ -245,6 +236,31 @@ func (s *GeminiMessagesCompatService) GetAntigravityGatewayService() *Antigravit
return s.antigravityGatewayService
}
func (s *GeminiMessagesCompatService) getSchedulableAccount(ctx context.Context, accountID int64) (*Account, error) {
if s.schedulerSnapshot != nil {
return s.schedulerSnapshot.GetAccount(ctx, accountID)
}
return s.accountRepo.GetByID(ctx, accountID)
}
func (s *GeminiMessagesCompatService) listSchedulableAccountsOnce(ctx context.Context, groupID *int64, platform string, hasForcePlatform bool) ([]Account, error) {
if s.schedulerSnapshot != nil {
accounts, _, err := s.schedulerSnapshot.ListSchedulableAccounts(ctx, groupID, platform, hasForcePlatform)
return accounts, err
}
useMixedScheduling := platform == PlatformGemini && !hasForcePlatform
queryPlatforms := []string{platform}
if useMixedScheduling {
queryPlatforms = []string{platform, PlatformAntigravity}
}
if groupID != nil {
return s.accountRepo.ListSchedulableByGroupIDAndPlatforms(ctx, *groupID, queryPlatforms)
}
return s.accountRepo.ListSchedulableByPlatforms(ctx, queryPlatforms)
}
func (s *GeminiMessagesCompatService) validateUpstreamBaseURL(raw string) (string, error) {
if s.cfg != nil && !s.cfg.Security.URLAllowlist.Enabled {
normalized, err := urlvalidator.ValidateURLFormat(raw, s.cfg.Security.URLAllowlist.AllowInsecureHTTP)
@@ -266,13 +282,7 @@ func (s *GeminiMessagesCompatService) validateUpstreamBaseURL(raw string) (strin
// HasAntigravityAccounts 检查是否有可用的 antigravity 账户
func (s *GeminiMessagesCompatService) HasAntigravityAccounts(ctx context.Context, groupID *int64) (bool, error) {
var accounts []Account
var err error
if groupID != nil {
accounts, err = s.accountRepo.ListSchedulableByGroupIDAndPlatform(ctx, *groupID, PlatformAntigravity)
} else {
accounts, err = s.accountRepo.ListSchedulableByPlatform(ctx, PlatformAntigravity)
}
accounts, err := s.listSchedulableAccountsOnce(ctx, groupID, PlatformAntigravity, false)
if err != nil {
return false, err
}
@@ -288,13 +298,7 @@ func (s *GeminiMessagesCompatService) HasAntigravityAccounts(ctx context.Context
// 3) OAuth accounts explicitly marked as ai_studio
// 4) Any remaining Gemini accounts (fallback)
func (s *GeminiMessagesCompatService) SelectAccountForAIStudioEndpoints(ctx context.Context, groupID *int64) (*Account, error) {
var accounts []Account
var err error
if groupID != nil {
accounts, err = s.accountRepo.ListSchedulableByGroupIDAndPlatform(ctx, *groupID, PlatformGemini)
} else {
accounts, err = s.accountRepo.ListSchedulableByPlatform(ctx, PlatformGemini)
}
accounts, err := s.listSchedulableAccountsOnce(ctx, groupID, PlatformGemini, true)
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}