Files
new-api-custom/service/group_degradation_test.go
T

235 lines
12 KiB
Go

package service
import (
"fmt"
"os"
"testing"
"time"
"github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/constant"
"github.com/QuantumNous/new-api/model"
fingerprint "github.com/QuantumNous/new-api/pkg/model_fingerprint"
"github.com/QuantumNous/new-api/setting/operation_setting"
"github.com/gin-gonic/gin"
"github.com/glebarez/sqlite"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
"gorm.io/driver/mysql"
"gorm.io/driver/postgres"
"gorm.io/gorm"
"gorm.io/gorm/schema"
"net/http/httptest"
)
func TestGroupDegradationDatabaseMatrix(t *testing.T) {
for _, dialect := range []string{"sqlite", "mysql", "postgres"} {
t.Run(dialect, func(t *testing.T) {
var driver gorm.Dialector
switch dialect {
case "sqlite":
driver = sqlite.Open(":memory:")
case "mysql":
if os.Getenv("TEST_MYSQL_DSN") == "" {
t.Skip("TEST_MYSQL_DSN unset")
}
driver = mysql.Open(os.Getenv("TEST_MYSQL_DSN"))
case "postgres":
if os.Getenv("TEST_POSTGRES_DSN") == "" {
t.Skip("TEST_POSTGRES_DSN unset")
}
driver = postgres.Open(os.Getenv("TEST_POSTGRES_DSN"))
}
db, err := gorm.Open(driver, &gorm.Config{NamingStrategy: schema.NamingStrategy{TablePrefix: fmt.Sprintf("degradation_%d_", time.Now().UnixNano())}})
require.NoError(t, err)
oldDB := model.DB
oldType, oldLogType := common.MainDatabaseType(), common.LogDatabaseType()
model.DB = db
common.SetDatabaseTypes(common.DatabaseType(dialect), oldLogType)
rows := []any{&model.Option{}, &model.Channel{}, &model.SystemTask{}, &model.DetectionSchedule{}, &model.DetectionRun{}, &model.DetectionCall{}, &model.Token{}}
require.NoError(t, db.AutoMigrate(rows...))
t.Cleanup(func() {
for _, row := range rows {
require.NoError(t, db.Migrator().DropTable(row))
}
model.DB = oldDB
common.SetDatabaseTypes(oldType, oldLogType)
groupDegradationSnapshot.updated = time.Time{}
groupDegradationSnapshot.latest = nil
conn, err := db.DB()
require.NoError(t, err)
require.NoError(t, conn.Close())
})
var version string
query := "select version()"
if dialect == "sqlite" {
query = "select sqlite_version()"
}
require.NoError(t, db.Raw(query).Scan(&version).Error)
t.Log("database version:", version)
now := time.Now().Unix()
require.NoError(t, db.Create(&model.Channel{Id: 1, Group: "default", Models: fingerprint.Model, Status: 1, Key: "test-only"}).Error)
body, err := common.Marshal(AstraSamplingState{Rounds: []AstraSamplingRound{{ID: "mismatch", ChannelID: 1, StartedAt: now - 30, FinishedAt: now - 10, Result: fingerprint.Result{Engine: fingerprint.Version, Status: "mismatch", Closest: "gpt-5.5", ValidSamples: 2}}}})
require.NoError(t, err)
require.NoError(t, db.Create(&model.SystemTask{TaskID: "degradation-test", Type: AstraSamplingTaskType, Status: model.SystemTaskStatusSucceeded, CreatedAt: now - 30, State: string(body)}).Error)
policy := &model.TokenRoutingPolicy{Version: 1, Mode: "custom", SkipDegradedGroups: true, Groups: []model.TokenGroupEntry{{Group: "default", Priority: 0}}}
encoded, err := common.Marshal(policy)
require.NoError(t, err)
token := model.Token{UserId: 1, Key: "degradation-test-key", RoutingPolicy: string(encoded)}
require.NoError(t, db.Create(&token).Error)
require.NoError(t, db.First(&token, token.Id).Error)
stored, err := token.GetRoutingPolicy()
require.NoError(t, err)
require.True(t, stored.SkipDegradedGroups)
ctx := func() *gin.Context {
c, _ := gin.CreateTestContext(httptest.NewRecorder())
c.Request = httptest.NewRequest("POST", "/v1/responses", nil)
c.Set(model.TokenRoutingContextKey, stored)
return c
}
groupDegradationSnapshot.updated = time.Time{}
assert.True(t, requestGroupDegraded(ctx(), fingerprint.Model, "default"))
smart := ctx()
smartPolicy := &model.TokenRoutingPolicy{Version: 1, Mode: "smart"}
smart.Set(model.TokenRoutingContextKey, smartPolicy)
assert.True(t, requestGroupDegraded(smart, fingerprint.Model, "default"))
smartPolicy.AllowDegraded = true
assert.False(t, requestGroupDegraded(smart, fingerprint.Model, "default"))
smartPolicy.AllowDegraded = false
assert.False(t, requestGroupDegraded(smart, "unmeasured-model", "default"))
manualResult, err := common.Marshal(fingerprint.Result{Engine: fingerprint.Version, Status: "match", Closest: fingerprint.Model, ValidSamples: 2})
require.NoError(t, err)
manual := model.DetectionRun{ID: "manual-recovery", RequestKey: "manual-recovery", ChannelID: 1, Kind: "fingerprint", UserID: 1, CreatedAt: now, Settled: true, Status: "succeeded", Result: string(manualResult)}
require.NoError(t, db.Create(&manual).Error)
groupDegradationSnapshot.updated = time.Time{}
assert.False(t, requestGroupDegraded(ctx(), fingerprint.Model, "default"))
pelican := model.DetectionRun{ID: "pelican-degraded", RequestKey: "pelican-degraded", ChannelID: 1, Kind: "pelican", CreatedAt: now, Settled: true, Status: "succeeded", Quality: "degraded"}
require.NoError(t, db.Create(&pelican).Error)
groupDegradationSnapshot.updated = time.Time{}
assert.False(t, requestGroupDegraded(ctx(), fingerprint.Model, "default"))
require.NoError(t, db.Model(&pelican).Update("quality", "normal").Error)
groupDegradationSnapshot.updated = time.Time{}
assert.False(t, requestGroupDegraded(ctx(), fingerprint.Model, "default"))
require.NoError(t, db.Delete(&manual).Error)
require.NoError(t, model.SaveStatusPageGroup(ctx().Request.Context(), "default", nil))
groupDegradationSnapshot.updated = time.Time{}
assert.False(t, requestGroupDegraded(ctx(), fingerprint.Model, "default"))
})
}
}
func TestGroupDegradationUsesLatestValidBeforeHistoryTruncation(t *testing.T) {
now := time.Now()
channels := []model.Channel{{Id: 1, Group: "default", Models: fingerprint.Model, Status: 1}}
rounds := []AstraSamplingRound{{ID: "valid", ChannelID: 1, StartedAt: now.Unix() - 100, FinishedAt: now.Unix() - 90, Result: fingerprint.Result{Engine: fingerprint.Version, Status: "mismatch", Closest: "gpt-5.5", ValidSamples: 1}}}
for i := range 15 {
rounds = append(rounds, AstraSamplingRound{ChannelID: 1, StartedAt: now.Unix() - int64(i), FinishedAt: now.Unix(), Result: fingerprint.Result{Engine: fingerprint.Version, Status: "error"}})
}
body, err := common.Marshal(AstraSamplingState{Rounds: rounds})
require.NoError(t, err)
tasks := []*model.SystemTask{{Status: model.SystemTaskStatusSucceeded, State: string(body)}}
summary := AggregateAstraSampling(channels, tasks, now)["default"]
require.NotNil(t, summary.LatestValid)
assert.Equal(t, "gpt-5.5", summary.LatestValid.Closest)
assert.Len(t, summary.History, 12)
rounds = append(rounds, AstraSamplingRound{ID: "recovered", ChannelID: 1, StartedAt: now.Unix(), FinishedAt: now.Unix(), Result: fingerprint.Result{Engine: fingerprint.Version, Status: "match", Closest: fingerprint.Model, ValidSamples: 2}})
body, err = common.Marshal(AstraSamplingState{Rounds: rounds})
require.NoError(t, err)
tasks[0].State = string(body)
summary = AggregateAstraSampling(channels, tasks, now)["default"]
require.NotNil(t, summary.LatestValid)
assert.Equal(t, fingerprint.Model, summary.LatestValid.Closest)
reset := aggregateAstraSampling(channels, tasks, now, map[string]model.StatusPageGroup{"default": {ResetBefore: now.Unix()}})
assert.Nil(t, reset["default"].LatestValid)
}
func TestGroupDegradationOverridesExistingPoolAffinity(t *testing.T) {
db := setupChannelSelectAutoGroupsTest(t)
createChannelSelectAutoGroupsChannel(t, db, 6401, "default", fingerprint.Model)
createChannelSelectAutoGroupsChannel(t, db, 6402, "vip", fingerprint.Model)
model.InitChannelCache()
ResetGroupAvailability()
t.Cleanup(ResetGroupAvailability)
previousRedis := common.RedisEnabled
common.RedisEnabled = false
t.Cleanup(func() { common.RedisEnabled = previousRedis; _ = poolBindings().Purge() })
require.NoError(t, poolBindings().Purge())
config := operation_setting.GetChannelAffinitySetting()
previous := *config
*config = operation_setting.ChannelAffinitySetting{Enabled: true, DefaultTTLSeconds: 3600, SessionMode: "strict", Rules: []operation_setting.ChannelAffinityRule{{Name: "degradation-test", ModelRegex: []string{".*"}, KeySources: []operation_setting.ChannelAffinityKeySource{{Type: "request_header", Key: "Session-Id"}}, SessionMode: "strict"}}}
t.Cleanup(func() { *config = previous })
ctx := func(degraded bool) *gin.Context {
c := setupGroupPoolContext(t)
c.Request = httptest.NewRequest("POST", "/v1/responses", nil)
c.Request.Header.Set("Session-Id", "degradation-session")
c.Set("id", 100)
c.Set("token_id", 100)
c.Set(model.TokenRoutingContextKey, &model.TokenRoutingPolicy{Version: 1, Mode: "custom", SkipDegradedGroups: true, Groups: []model.TokenGroupEntry{{Group: "default", Priority: 0}, {Group: "vip", Priority: 1}}})
rounds := map[string]AstraSamplingHistory{}
if degraded {
rounds["default"] = AstraSamplingHistory{At: time.Now().Unix(), Closest: "gpt-5.5", ValidSamples: 1}
}
c.Set("pool_group_fingerprints", rounds)
return c
}
first := ctx(false)
selected, group, err := SelectChannelForRequest(first, fingerprint.Model, &RetryParam{Ctx: first, TokenGroup: "default", ModelName: fingerprint.Model})
require.Nil(t, err)
require.NotNil(t, selected)
require.Equal(t, 6401, selected.Id)
first.Set("channel_id", selected.Id)
common.SetContextKey(first, constant.ContextKeyAutoGroup, group)
RequestPolicy(first).Successful = true
RecordPoolAffinity(first)
bound, _ := PreferredPoolChannel(ctx(false), fingerprint.Model)
require.NotNil(t, bound)
next := ctx(true)
selected, group, err = SelectChannelForRequest(next, fingerprint.Model, &RetryParam{Ctx: next, TokenGroup: "default", ModelName: fingerprint.Model})
require.Nil(t, err)
require.NotNil(t, selected)
assert.Equal(t, 6402, selected.Id)
assert.Equal(t, "vip", group)
}
func TestGroupDegradationSkipsPoolGroupsAndRespectsUnknownRecoveryAndOptOut(t *testing.T) {
db := setupChannelSelectAutoGroupsTest(t)
createChannelSelectAutoGroupsChannel(t, db, 6301, "default", fingerprint.Model)
createChannelSelectAutoGroupsChannel(t, db, 6302, "vip", fingerprint.Model)
model.InitChannelCache()
ResetGroupAvailability()
t.Cleanup(ResetGroupAvailability)
for _, test := range []struct {
name string
enabled bool
closest string
age time.Duration
degradeBackup bool
want int
}{
{"latest mismatch switches", true, "gpt-5.5", time.Minute, false, 6302},
{"recovered group returns", true, fingerprint.Model, time.Minute, false, 6301},
{"unknown stays eligible", true, "", time.Minute, false, 6301},
{"valid result persists beyond window", true, "gpt-5.5", 7 * time.Hour, false, 6302},
{"disabled keeps routing", false, "gpt-5.5", time.Minute, false, 6301},
{"all degraded has no route", true, "gpt-5.5", time.Minute, true, 0},
} {
t.Run(test.name, func(t *testing.T) {
c := setupGroupPoolContext(t)
c.Set(model.TokenRoutingContextKey, &model.TokenRoutingPolicy{Version: 1, Mode: "custom", SkipDegradedGroups: test.enabled, Groups: []model.TokenGroupEntry{{Group: "default", Priority: 0}, {Group: "vip", Priority: 1}}})
rounds := map[string]AstraSamplingHistory{"default": {At: time.Now().Add(-test.age).Unix(), Closest: test.closest, ValidSamples: 1}}
if test.degradeBackup {
rounds["vip"] = rounds["default"]
}
c.Set("pool_group_fingerprints", rounds)
selected, _, err := selectTokenRoutingPool(&RetryParam{Ctx: c, TokenGroup: "default", ModelName: fingerprint.Model})
require.NoError(t, err)
if test.want == 0 {
assert.Nil(t, selected)
return
}
require.NotNil(t, selected)
assert.Equal(t, test.want, selected.Id)
})
}
}