feat(cache,worker): overhaul cache layer and add generation task lease recovery
Deployment Config CI / Deployment Config (push) Successful in 24s
Backend CI / Backend (push) Successful in 14m33s

- shared/cache: add Options/L1/async/metrics/prefix decorators, multi-key ops, Redis pool tuning, and JSON readthrough metrics
- worker-generate: claim tasks via DB lease + heartbeat, requeue stale queued tasks, expire dead leases with refund/cache invalidation
- tenant: version article cache keys so worker recovery invalidations propagate cleanly
- shared/config: expand Redis (pool/timeouts/TLS) and Generation (lease/recovery) configs with defaults

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-03 02:02:39 +08:00
parent bbfeabdaa5
commit c1ef717d17
37 changed files with 2502 additions and 93 deletions
@@ -2,7 +2,6 @@ package generate
import (
"context"
"database/sql"
"encoding/json"
"errors"
"fmt"
@@ -10,11 +9,11 @@ import (
"strings"
"time"
"github.com/jackc/pgx/v5"
"github.com/jackc/pgx/v5/pgxpool"
amqp "github.com/rabbitmq/amqp091-go"
"go.uber.org/zap"
sharedcache "github.com/geo-platform/tenant-api/internal/shared/cache"
"github.com/geo-platform/tenant-api/internal/shared/config"
"github.com/geo-platform/tenant-api/internal/shared/messaging/rabbitmq"
tenantapp "github.com/geo-platform/tenant-api/internal/tenant/app"
@@ -31,9 +30,11 @@ type ArticleGenerationWorker struct {
promptRuleService *tenantapp.PromptRuleGenerationService
imitationService *tenantapp.ArticleImitationService
kolWorker *KolGenerationWorker
cache sharedcache.Cache
logger *zap.Logger
retryInterval time.Duration
processTimeout time.Duration
generationCfg config.GenerationConfig
configProvider config.Provider
consumerPrefix string
workerConcurrency int
@@ -64,6 +65,7 @@ func NewArticleGenerationWorker(
logger: logger,
retryInterval: defaultArticleGenerationWorkerRetryInterval,
processTimeout: articleGenerationProcessTimeout(cfg),
generationCfg: cfg,
consumerPrefix: fmt.Sprintf("worker-generate-%d", os.Getpid()),
workerConcurrency: workerConcurrency,
}
@@ -74,6 +76,11 @@ func (w *ArticleGenerationWorker) WithConfigProvider(provider config.Provider) *
return w
}
func (w *ArticleGenerationWorker) WithCache(c sharedcache.Cache) *ArticleGenerationWorker {
w.cache = c
return w
}
func (w *ArticleGenerationWorker) Start(ctx context.Context) {
if w == nil || w.pool == nil || w.rabbitMQ == nil {
return
@@ -83,6 +90,7 @@ func (w *ArticleGenerationWorker) Start(ctx context.Context) {
consumerName := fmt.Sprintf("%s-%d", w.consumerPrefix, i+1)
go w.run(ctx, consumerName)
}
w.startRecovery(ctx)
}
func (w *ArticleGenerationWorker) run(ctx context.Context, consumerName string) {
@@ -135,31 +143,25 @@ func (w *ArticleGenerationWorker) handleDelivery(parent context.Context, deliver
return
}
recoveryCfg := w.runtimeRecoveryConfig()
ctx, cancel := context.WithTimeout(parent, w.runtimeProcessTimeout())
defer cancel()
task, terminal, err := w.loadTask(ctx, envelope.TaskID)
lease, terminal, err := w.claimGenerationTask(ctx, envelope.TaskID, generationWorkerOwner(delivery.ConsumerTag), recoveryCfg.LeaseTTL)
if err != nil {
if errors.Is(err, pgx.ErrNoRows) {
if ackErr := delivery.Ack(false); ackErr != nil && w.logger != nil {
w.logger.Warn("article generation ack failed for missing task",
zap.Error(ackErr),
zap.Int64("task_id", envelope.TaskID),
)
}
return
}
if task.ID > 0 {
if lease != nil {
task := lease.task.toClaimedTask()
w.handleTaskFailure(ctx, task, "task_load", err)
w.releaseGenerationTaskLease(context.Background(), task.ID, lease.token)
if ackErr := delivery.Ack(false); ackErr != nil && w.logger != nil {
w.logger.Warn("article generation ack failed after task load failure",
w.logger.Warn("article generation ack failed after task claim failure",
zap.Error(ackErr),
zap.Int64("task_id", envelope.TaskID),
)
}
return
}
requeue := !delivery.Redelivered
requeue := !delivery.Redelivered && !isPermanentGenerationTaskDecodeError(err)
w.rejectDelivery(delivery, requeue, err, envelope.TaskID)
return
}
@@ -172,6 +174,18 @@ func (w *ArticleGenerationWorker) handleDelivery(parent context.Context, deliver
}
return
}
if lease == nil {
w.rejectDelivery(delivery, !delivery.Redelivered, fmt.Errorf("generation task claim returned empty lease"), envelope.TaskID)
return
}
task := lease.task.toClaimedTask()
heartbeatCtx, stopHeartbeat := context.WithCancel(parent)
go w.runGenerationLeaseHeartbeat(heartbeatCtx, task.ID, lease.token, recoveryCfg.LeaseTTL)
defer func() {
stopHeartbeat()
w.releaseGenerationTaskLease(context.Background(), task.ID, lease.token)
}()
task.TaskType = strings.TrimSpace(task.TaskType)
switch {
@@ -215,10 +229,13 @@ func (w *ArticleGenerationWorker) runtimeProcessTimeout() time.Duration {
if w != nil && w.processTimeout > 0 {
return w.processTimeout
}
return articleGenerationProcessTimeout(config.GenerationConfig{})
cfg := config.GenerationConfig{}
config.NormalizeGenerationConfig(&cfg)
return articleGenerationProcessTimeout(cfg)
}
func articleGenerationProcessTimeout(cfg config.GenerationConfig) time.Duration {
config.NormalizeGenerationConfig(&cfg)
processTimeout := cfg.ArticleTimeout + 2*time.Minute
if processTimeout <= 2*time.Minute {
return 10 * time.Minute
@@ -238,50 +255,11 @@ func (w *ArticleGenerationWorker) handleTaskFailure(ctx context.Context, task te
}
}
func (w *ArticleGenerationWorker) loadTask(ctx context.Context, taskID int64) (tenantapp.ClaimedGenerationTask, bool, error) {
var (
task tenantapp.ClaimedGenerationTask
status string
operatorID sql.NullInt64
articleID sql.NullInt64
reservationID sql.NullInt64
inputParamsJSON []byte
)
err := w.pool.QueryRow(ctx, `
SELECT id, tenant_id, operator_id, article_id, quota_reservation_id, task_type, input_params_json, status
FROM generation_tasks
WHERE id = $1
`, taskID).Scan(
&task.ID,
&task.TenantID,
&operatorID,
&articleID,
&reservationID,
&task.TaskType,
&inputParamsJSON,
&status,
)
if err != nil {
return tenantapp.ClaimedGenerationTask{}, false, err
func isPermanentGenerationTaskDecodeError(err error) bool {
if err == nil {
return false
}
if status == "completed" || status == "failed" {
return tenantapp.ClaimedGenerationTask{}, true, nil
}
task.OperatorID = operatorID.Int64
task.ArticleID = articleID.Int64
task.QuotaReservationID = reservationID.Int64
params, err := parseJSONMap(inputParamsJSON)
if err != nil {
task.InputParams = map[string]interface{}{}
return task, false, fmt.Errorf("decode generation input params: %w", err)
}
task.InputParams = params
return task, false, nil
return strings.Contains(err.Error(), "decode generation input params")
}
func (w *ArticleGenerationWorker) rejectDelivery(delivery amqp.Delivery, requeue bool, err error, taskID int64) {