Skip to content

fix(scheduler): preserve pre-hydration account metadata #5657

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
lyy0709 wants to merge 1 commit into
base: main
Choose a base branch
from
+181 −15
Open

fix(scheduler): preserve pre-hydration account metadata #5657

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
69 changes: 54 additions & 15 deletions backend/internal/repository/scheduler_cache.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,18 +15,22 @@ import (
)

const (
schedulerBucketSetKey = "sched:buckets"
schedulerOutboxWatermarkKey = "sched:outbox:watermark"
schedulerAccountPrefix = "sched:acc:"
schedulerAccountMetaPrefix = "sched:meta:"
schedulerAccountLastUsedPrefix = "sched:acc:last_used:"
schedulerActivePrefix = "sched:active:"
schedulerReadyPrefix = "sched:ready:"
schedulerVersionPrefix = "sched:ver:"
schedulerEpochPrefix = "sched:epoch:"
schedulerRetiredPrefix = "sched:retired:"
schedulerSnapshotPrefix = "sched:"
schedulerLockPrefix = "sched:lock:"
schedulerBucketSetKey = "sched:buckets"
schedulerOutboxWatermarkKey = "sched:outbox:watermark"
schedulerAccountPrefix = "sched:acc:"
// schedulerAccountMetaPrefix is schema-versioned because metadata is read
// before the selected account is hydrated. Bump the namespace whenever a
// projection change affects pre-hydration scheduling behavior.
schedulerAccountMetaPrefix = "sched:meta:v2:"
schedulerLegacyAccountMetaPrefix = "sched:meta:"
schedulerAccountLastUsedPrefix = "sched:acc:last_used:"
schedulerActivePrefix = "sched:active:"
schedulerReadyPrefix = "sched:ready:"
schedulerVersionPrefix = "sched:ver:"
schedulerEpochPrefix = "sched:epoch:"
schedulerRetiredPrefix = "sched:retired:"
schedulerSnapshotPrefix = "sched:"
schedulerLockPrefix = "sched:lock:"

defaultSchedulerSnapshotMGetChunkSize = 128
defaultSchedulerSnapshotWriteChunkSize = 256
Expand Down Expand Up @@ -604,7 +608,12 @@ func (c *schedulerCache) DeleteAccount(ctx context.Context, accountID int64) err
return nil
}
id := strconv.FormatInt(accountID, 10)
return c.rdb.Del(ctx, schedulerAccountKey(id), schedulerAccountMetaKey(id), schedulerLastUsedKey(id)).Err()
return c.rdb.Del(ctx,
schedulerAccountKey(id),
schedulerAccountMetaKey(id),
schedulerLegacyAccountMetaKey(id),
schedulerLastUsedKey(id),
).Err()
}

func (c *schedulerCache) UpdateLastUsed(ctx context.Context, updates map[int64]time.Time) error {
Expand Down Expand Up @@ -636,7 +645,12 @@ func (c *schedulerCache) UpdateLastUsed(ctx context.Context, updates map[int64]t
"error", err,
)
idText := strconv.FormatInt(id, 10)
pipe.Del(ctx, schedulerAccountKey(idText), schedulerAccountMetaKey(idText), schedulerLastUsedKey(idText))
pipe.Del(ctx,
schedulerAccountKey(idText),
schedulerAccountMetaKey(idText),
schedulerLegacyAccountMetaKey(idText),
schedulerLastUsedKey(idText),
)
queued++
continue
}
Expand Down Expand Up @@ -720,6 +734,10 @@ func schedulerAccountMetaKey(id string) string {
return schedulerAccountMetaPrefix + id
}

func schedulerLegacyAccountMetaKey(id string) string {
return schedulerLegacyAccountMetaPrefix + id
}

func schedulerLastUsedKey(id string) string {
return schedulerAccountLastUsedPrefix + id
}
Expand Down Expand Up @@ -809,6 +827,7 @@ func (c *schedulerCache) writeAccountIDs(ctx context.Context, accounts []service
id := strconv.FormatInt(account.ID, 10)
pipe.Set(ctx, schedulerAccountKey(id), fullPayload, 0)
pipe.Set(ctx, schedulerAccountMetaKey(id), metaPayload, 0)
pipe.Del(ctx, schedulerLegacyAccountMetaKey(id))
// Keep the hot LastUsedAt side key untouched: a lagging snapshot rebuild
// must not overwrite a newer scheduler update.
accountIDs = append(accountIDs, account.ID)
Expand Down Expand Up @@ -954,7 +973,17 @@ func filterSchedulerCredentials(credentials map[string]any) map[string]any {
if len(credentials) == 0 {
return nil
}
keys := []string{"model_mapping", "compact_model_mapping", "api_key", "project_id", "oauth_type", "plan_type"}
keys := []string{
"model_mapping",
"compact_model_mapping",
"api_key",
"project_id",
"oauth_type",
"plan_type",
"account_scheduling_threshold",
"subscription_tier",
"team_id",
}
filtered := make(map[string]any)
for _, key := range keys {
if value, ok := credentials[key]; ok && value != nil {
Expand Down Expand Up @@ -1000,6 +1029,16 @@ func filterSchedulerExtra(extra map[string]any) map[string]any {
"openai_ws_force_http",
"openai_responses_mode",
"openai_responses_supported",
"openai_passthrough",
"openai_oauth_passthrough",
"openai_compact_mode",
"openai_compact_supported",
"privacy_mode",
"session_window_utilization",
"passive_usage_7d_utilization",
"passive_usage_7d_reset",
"grok_sched_utilization",
"grok_sched_reset_at",
"codex_5h_used_percent",
"codex_7d_used_percent",
"codex_5h_reset_at",
Expand Down
127 changes: 127 additions & 0 deletions backend/internal/repository/scheduler_cache_unit_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,13 +75,16 @@ func TestSchedulerCacheUpdateLastUsedClearsUnencodablePayload(t *testing.T) {
cache := newSchedulerCacheUnit(t)
account := service.Account{ID: 114, Platform: service.PlatformOpenAI, Type: service.AccountTypeAPIKey}
require.NoError(t, cache.SetAccount(ctx, &account))
legacyKey := schedulerLegacyAccountMetaKey(strconv.FormatInt(account.ID, 10))
require.NoError(t, cache.rdb.Set(ctx, legacyKey, `{}`, 0).Err())

invalidTime := time.Date(10000, time.January, 1, 0, 0, 0, 0, time.UTC)
require.NoError(t, cache.UpdateLastUsed(ctx, map[int64]time.Time{account.ID: invalidTime}))

cached, err := cache.GetAccount(ctx, account.ID)
require.NoError(t, err)
require.Nil(t, cached)
require.EqualValues(t, 0, cache.rdb.Exists(ctx, legacyKey).Val())
}

func TestSchedulerCacheSnapshotAccountIDReusePreservesPayloadAndMembers(t *testing.T) {
Expand Down Expand Up @@ -369,6 +372,130 @@ func TestBuildSchedulerMetadataAccount_KeepsGrokMediaEligibility(t *testing.T) {
})
}

func TestBuildSchedulerMetadataAccount_KeepsPreHydrationSchedulingFields(t *testing.T) {
now := time.Now().UTC().Truncate(time.Second)

t.Run("anthropic threshold override and windows", func(t *testing.T) {
windowEnd := now.Add(2 * time.Hour)
account := service.Account{
ID: 90,
Platform: service.PlatformAnthropic,
Type: service.AccountTypeOAuth,
Status: service.StatusActive,
Schedulable: true,
SessionWindowEnd: &windowEnd,
Credentials: map[string]any{
"account_scheduling_threshold": 80,
"access_token": "must-not-enter-metadata",
},
Extra: map[string]any{
"session_window_utilization": 0.85,
"passive_usage_7d_utilization": 0.90,
"passive_usage_7d_reset": now.Add(7 * 24 * time.Hour).Format(time.RFC3339),
"unrelated": "drop-me",
},
}

got := buildSchedulerMetadataAccount(account)
decision := service.EvaluateAccountSchedulingThreshold(&got, map[string]int{
service.PlatformAnthropic: 95,
}, now)

require.True(t, decision.ShouldPause)
require.Equal(t, 80, decision.ThresholdPercent)
require.Equal(t, "7d", decision.Window)
require.Empty(t, got.GetCredential("access_token"))
require.NotContains(t, got.Extra, "unrelated")
})

t.Run("grok threshold and scheduling identity", func(t *testing.T) {
account := service.Account{
ID: 91,
Platform: service.PlatformGrok,
Type: service.AccountTypeOAuth,
Status: service.StatusActive,
Schedulable: true,
Credentials: map[string]any{
"account_scheduling_threshold": 80,
"subscription_tier": "free",
"team_id": "team-a",
"refresh_token": "must-not-enter-metadata",
},
Extra: map[string]any{
"grok_sched_utilization": 90.0,
"grok_sched_reset_at": now.Add(time.Hour).Format(time.RFC3339),
},
}

got := buildSchedulerMetadataAccount(account)
decision := service.EvaluateAccountSchedulingThreshold(&got, map[string]int{
service.PlatformGrok: 95,
}, now)

require.True(t, decision.ShouldPause)
require.Equal(t, 80, decision.ThresholdPercent)
require.Equal(t, "quota", decision.Window)
require.Equal(t, "free", got.GetCredential("subscription_tier"))
require.Equal(t, "team-a", got.GetCredential("team_id"))
require.Empty(t, got.GetCredential("refresh_token"))
})

t.Run("openai privacy passthrough and compact routing", func(t *testing.T) {
account := service.Account{
ID: 92,
Platform: service.PlatformOpenAI,
Type: service.AccountTypeOAuth,
Credentials: map[string]any{
"model_mapping": map[string]any{"known-model": "known-model"},
},
Extra: map[string]any{
"privacy_mode": service.PrivacyModeTrainingOff,
"openai_passthrough": true,
"openai_compact_mode": service.OpenAICompactModeForceOn,
"openai_compact_supported": true,
},
}

got := buildSchedulerMetadataAccount(account)
supported, known := got.OpenAICompactSupportKnown()

require.True(t, got.IsPrivacySet())
require.True(t, got.IsOpenAIPassthroughEnabled())
require.True(t, got.IsModelSupported("unmapped-upstream-model"))
require.True(t, known)
require.True(t, supported)
})
}

func TestSchedulerMetadataSchemaVersionRejectsAndCleansLegacyPayload(t *testing.T) {
ctx := context.Background()
cache := newSchedulerCacheUnit(t)
account := service.Account{ID: 93, Platform: service.PlatformGrok, Type: service.AccountTypeOAuth}
bucket := service.SchedulerBucket{GroupID: 93, Platform: service.PlatformGrok, Mode: service.SchedulerModeSingle}
token, err := cache.CaptureBucketWriteToken(ctx, bucket)
require.NoError(t, err)
require.NoError(t, cache.SetSnapshot(ctx, bucket, token, []service.Account{account}))

id := strconv.FormatInt(account.ID, 10)
metadata, err := cache.rdb.Get(ctx, schedulerAccountMetaKey(id)).Bytes()
require.NoError(t, err)
require.NoError(t, cache.rdb.Del(ctx, schedulerAccountMetaKey(id)).Err())
require.NoError(t, cache.rdb.Set(ctx, schedulerLegacyAccountMetaKey(id), metadata, 0).Err())

snapshot, hit, err := cache.GetSnapshot(ctx, bucket)
require.NoError(t, err)
require.False(t, hit, "legacy metadata must force a database rebuild")
require.Nil(t, snapshot)

require.NoError(t, cache.SetAccount(ctx, &account))
require.EqualValues(t, 0, cache.rdb.Exists(ctx, schedulerLegacyAccountMetaKey(id)).Val())
require.EqualValues(t, 1, cache.rdb.Exists(ctx, schedulerAccountMetaKey(id)).Val())

require.NoError(t, cache.rdb.Set(ctx, schedulerLegacyAccountMetaKey(id), metadata, 0).Err())
require.NoError(t, cache.DeleteAccount(ctx, account.ID))
require.EqualValues(t, 0, cache.rdb.Exists(ctx, schedulerLegacyAccountMetaKey(id)).Val())
}

func TestBuildSchedulerMetadataAccount_KeepsSlimGroupMembership(t *testing.T) {
account := service.Account{
ID: 42,
Expand Down
Loading