2
0
mirror of https://github.com/hibiken/asynq.git synced 2026-08-15 21:08:00 +08:00
Files
asynq/client.go
Erik Nilsen 06a06970d6 Document atomicity guarantees and fix NOSCRIPT bug in BatchEnqueue
Add comprehensive doc comments to BatchEnqueueContext, the Broker
interface method, and the RDB implementation explaining that the batch
uses a Redis pipeline (not MULTI/EXEC), so partial success is possible
and individual Lua scripts are atomic but the batch is not.

Fix a bug where Script.Run inside a pipeline only sends EVALSHA without
the automatic EVAL fallback that non-pipeline calls get. On a fresh
Redis (or after SCRIPT FLUSH), this caused NOSCRIPT errors for every
pipeline-batched script invocation. The fix preloads the required Lua
scripts before building the pipeline.

Also roll back the in-memory queuesPublished cache when the pipeline
fails, preventing stale entries from suppressing future SADD calls.
2026-04-09 11:35:11 -07:00

585 lines
19 KiB
Go

// Copyright 2020 Kentaro Hibino. All rights reserved.
// Use of this source code is governed by a MIT license
// that can be found in the LICENSE file.
package asynq
import (
"context"
"fmt"
"strings"
"time"
"github.com/google/uuid"
"github.com/hibiken/asynq/internal/base"
"github.com/hibiken/asynq/internal/errors"
"github.com/hibiken/asynq/internal/rdb"
"github.com/redis/go-redis/v9"
)
// A Client is responsible for scheduling tasks.
//
// A Client is used to register tasks that should be processed
// immediately or some time in the future.
//
// Clients are safe for concurrent use by multiple goroutines.
type Client struct {
broker base.Broker
// When a Client has been created with an existing Redis connection, we do
// not want to close it.
sharedConnection bool
}
// NewClient returns a new Client instance given a redis connection option.
func NewClient(r RedisConnOpt) *Client {
redisClient, ok := r.MakeRedisClient().(redis.UniversalClient)
if !ok {
panic(fmt.Sprintf("asynq: unsupported RedisConnOpt type %T", r))
}
client := NewClientFromRedisClient(redisClient)
client.sharedConnection = false
return client
}
// NewClientFromRedisClient returns a new instance of Client given a redis.UniversalClient
// Warning: The underlying redis connection pool will not be closed by Asynq, you are responsible for closing it.
func NewClientFromRedisClient(c redis.UniversalClient) *Client {
return &Client{broker: rdb.NewRDB(c), sharedConnection: true}
}
type OptionType int
const (
MaxRetryOpt OptionType = iota
QueueOpt
TimeoutOpt
DeadlineOpt
UniqueOpt
ProcessAtOpt
ProcessInOpt
TaskIDOpt
RetentionOpt
GroupOpt
)
// Option specifies the task processing behavior.
type Option interface {
// String returns a string representation of the option.
String() string
// Type describes the type of the option.
Type() OptionType
// Value returns a value used to create this option.
Value() interface{}
}
// Internal option representations.
type (
retryOption int
queueOption string
taskIDOption string
timeoutOption time.Duration
deadlineOption time.Time
uniqueOption time.Duration
processAtOption time.Time
processInOption time.Duration
retentionOption time.Duration
groupOption string
)
// MaxRetry returns an option to specify the max number of times
// the task will be retried.
//
// Negative retry count is treated as zero retry.
func MaxRetry(n int) Option {
if n < 0 {
n = 0
}
return retryOption(n)
}
func (n retryOption) String() string { return fmt.Sprintf("MaxRetry(%d)", int(n)) }
func (n retryOption) Type() OptionType { return MaxRetryOpt }
func (n retryOption) Value() interface{} { return int(n) }
// Queue returns an option to specify the queue to enqueue the task into.
func Queue(name string) Option {
return queueOption(name)
}
func (name queueOption) String() string { return fmt.Sprintf("Queue(%q)", string(name)) }
func (name queueOption) Type() OptionType { return QueueOpt }
func (name queueOption) Value() interface{} { return string(name) }
// TaskID returns an option to specify the task ID.
func TaskID(id string) Option {
return taskIDOption(id)
}
func (id taskIDOption) String() string { return fmt.Sprintf("TaskID(%q)", string(id)) }
func (id taskIDOption) Type() OptionType { return TaskIDOpt }
func (id taskIDOption) Value() interface{} { return string(id) }
// Timeout returns an option to specify how long a task may run.
// If the timeout elapses before the Handler returns, then the task
// will be retried.
//
// Zero duration means no limit.
//
// If there's a conflicting Deadline option, whichever comes earliest
// will be used.
func Timeout(d time.Duration) Option {
return timeoutOption(d)
}
func (d timeoutOption) String() string { return fmt.Sprintf("Timeout(%v)", time.Duration(d)) }
func (d timeoutOption) Type() OptionType { return TimeoutOpt }
func (d timeoutOption) Value() interface{} { return time.Duration(d) }
// Deadline returns an option to specify the deadline for the given task.
// If it reaches the deadline before the Handler returns, then the task
// will be retried.
//
// If there's a conflicting Timeout option, whichever comes earliest
// will be used.
func Deadline(t time.Time) Option {
return deadlineOption(t)
}
func (t deadlineOption) String() string {
return fmt.Sprintf("Deadline(%v)", time.Time(t).Format(time.UnixDate))
}
func (t deadlineOption) Type() OptionType { return DeadlineOpt }
func (t deadlineOption) Value() interface{} { return time.Time(t) }
// Unique returns an option to enqueue a task only if the given task is unique.
// Task enqueued with this option is guaranteed to be unique within the given ttl.
// Once the task gets processed successfully or once the TTL has expired,
// another task with the same uniqueness may be enqueued.
// ErrDuplicateTask error is returned when enqueueing a duplicate task.
// TTL duration must be greater than or equal to 1 second.
//
// Uniqueness of a task is based on the following properties:
// - Task Type
// - Task Payload
// - Queue Name
func Unique(ttl time.Duration) Option {
return uniqueOption(ttl)
}
func (ttl uniqueOption) String() string { return fmt.Sprintf("Unique(%v)", time.Duration(ttl)) }
func (ttl uniqueOption) Type() OptionType { return UniqueOpt }
func (ttl uniqueOption) Value() interface{} { return time.Duration(ttl) }
// ProcessAt returns an option to specify when to process the given task.
//
// If there's a conflicting ProcessIn option, the last option passed to Enqueue overrides the others.
func ProcessAt(t time.Time) Option {
return processAtOption(t)
}
func (t processAtOption) String() string {
return fmt.Sprintf("ProcessAt(%v)", time.Time(t).Format(time.UnixDate))
}
func (t processAtOption) Type() OptionType { return ProcessAtOpt }
func (t processAtOption) Value() interface{} { return time.Time(t) }
// ProcessIn returns an option to specify when to process the given task relative to the current time.
//
// If there's a conflicting ProcessAt option, the last option passed to Enqueue overrides the others.
func ProcessIn(d time.Duration) Option {
return processInOption(d)
}
func (d processInOption) String() string { return fmt.Sprintf("ProcessIn(%v)", time.Duration(d)) }
func (d processInOption) Type() OptionType { return ProcessInOpt }
func (d processInOption) Value() interface{} { return time.Duration(d) }
// Retention returns an option to specify the duration of retention period for the task.
// If this option is provided, the task will be stored as a completed task after successful processing.
// A completed task will be deleted after the specified duration elapses.
func Retention(d time.Duration) Option {
return retentionOption(d)
}
func (ttl retentionOption) String() string { return fmt.Sprintf("Retention(%v)", time.Duration(ttl)) }
func (ttl retentionOption) Type() OptionType { return RetentionOpt }
func (ttl retentionOption) Value() interface{} { return time.Duration(ttl) }
// Group returns an option to specify the group used for the task.
// Tasks in a given queue with the same group will be aggregated into one task before passed to Handler.
func Group(name string) Option {
return groupOption(name)
}
func (name groupOption) String() string { return fmt.Sprintf("Group(%q)", string(name)) }
func (name groupOption) Type() OptionType { return GroupOpt }
func (name groupOption) Value() interface{} { return string(name) }
// ErrDuplicateTask indicates that the given task could not be enqueued since it's a duplicate of another task.
//
// ErrDuplicateTask error only applies to tasks enqueued with a Unique option.
var ErrDuplicateTask = errors.New("task already exists")
// ErrTaskIDConflict indicates that the given task could not be enqueued since its task ID already exists.
//
// ErrTaskIDConflict error only applies to tasks enqueued with a TaskID option.
var ErrTaskIDConflict = errors.New("task ID conflicts with another task")
type option struct {
retry int
queue string
taskID string
timeout time.Duration
deadline time.Time
uniqueTTL time.Duration
processAt time.Time
retention time.Duration
group string
}
// composeOptions merges user provided options into the default options
// and returns the composed option.
// It also validates the user provided options and returns an error if any of
// the user provided options fail the validations.
func composeOptions(opts ...Option) (option, error) {
res := option{
retry: defaultMaxRetry,
queue: base.DefaultQueueName,
taskID: uuid.NewString(),
timeout: 0, // do not set to defaultTimeout here
deadline: time.Time{},
processAt: time.Now(),
}
for _, opt := range opts {
switch opt := opt.(type) {
case retryOption:
res.retry = int(opt)
case queueOption:
qname := string(opt)
if err := base.ValidateQueueName(qname); err != nil {
return option{}, err
}
res.queue = qname
case taskIDOption:
id := string(opt)
if isBlank(id) {
return option{}, errors.New("task ID cannot be empty")
}
res.taskID = id
case timeoutOption:
res.timeout = time.Duration(opt)
case deadlineOption:
res.deadline = time.Time(opt)
case uniqueOption:
ttl := time.Duration(opt)
if ttl < 1*time.Second {
return option{}, errors.New("Unique TTL cannot be less than 1s")
}
res.uniqueTTL = ttl
case processAtOption:
res.processAt = time.Time(opt)
case processInOption:
res.processAt = time.Now().Add(time.Duration(opt))
case retentionOption:
res.retention = time.Duration(opt)
case groupOption:
key := string(opt)
if isBlank(key) {
return option{}, errors.New("group key cannot be empty")
}
res.group = key
default:
// ignore unexpected option
}
}
return res, nil
}
// isBlank returns true if the given s is empty or consist of all whitespaces.
func isBlank(s string) bool {
return strings.TrimSpace(s) == ""
}
const (
// Default max retry count used if nothing is specified.
defaultMaxRetry = 25
// Default timeout used if both timeout and deadline are not specified.
defaultTimeout = 30 * time.Minute
)
// Value zero indicates no timeout and no deadline.
var (
noTimeout time.Duration = 0
noDeadline time.Time = time.Unix(0, 0)
)
// Close closes the connection with redis.
func (c *Client) Close() error {
if c.sharedConnection {
return fmt.Errorf("redis connection is shared so the Client can't be closed through asynq")
}
return c.broker.Close()
}
// Enqueue enqueues the given task to a queue.
//
// Enqueue returns TaskInfo and nil error if the task is enqueued successfully, otherwise returns a non-nil error.
//
// The argument opts specifies the behavior of task processing.
// If there are conflicting Option values the last one overrides others.
// Any options provided to NewTask can be overridden by options passed to Enqueue.
// By default, max retry is set to 25 and timeout is set to 30 minutes.
//
// If no ProcessAt or ProcessIn options are provided, the task will be pending immediately.
//
// Enqueue uses context.Background internally; to specify the context, use EnqueueContext.
func (c *Client) Enqueue(task *Task, opts ...Option) (*TaskInfo, error) {
return c.EnqueueContext(context.Background(), task, opts...)
}
// EnqueueContext enqueues the given task to a queue.
//
// EnqueueContext returns TaskInfo and nil error if the task is enqueued successfully, otherwise returns a non-nil error.
//
// The argument opts specifies the behavior of task processing.
// If there are conflicting Option values the last one overrides others.
// Any options provided to NewTask can be overridden by options passed to Enqueue.
// By default, max retry is set to 25 and timeout is set to 30 minutes.
//
// If no ProcessAt or ProcessIn options are provided, the task will be pending immediately.
//
// The first argument context applies to the enqueue operation. To specify task timeout and deadline, use Timeout and Deadline option instead.
func (c *Client) EnqueueContext(ctx context.Context, task *Task, opts ...Option) (*TaskInfo, error) {
if task == nil {
return nil, fmt.Errorf("task cannot be nil")
}
if strings.TrimSpace(task.Type()) == "" {
return nil, fmt.Errorf("task typename cannot be empty")
}
// merge task options with the options provided at enqueue time.
opts = append(task.opts, opts...)
opt, err := composeOptions(opts...)
if err != nil {
return nil, err
}
deadline := noDeadline
if !opt.deadline.IsZero() {
deadline = opt.deadline
}
timeout := noTimeout
if opt.timeout != 0 {
timeout = opt.timeout
}
if deadline.Equal(noDeadline) && timeout == noTimeout {
// If neither deadline nor timeout are set, use default timeout.
timeout = defaultTimeout
}
var uniqueKey string
if opt.uniqueTTL > 0 {
uniqueKey = base.UniqueKey(opt.queue, task.Type(), task.Payload())
}
msg := &base.TaskMessage{
ID: opt.taskID,
Type: task.Type(),
Payload: task.Payload(),
Headers: task.Headers(),
Queue: opt.queue,
Retry: opt.retry,
Deadline: deadline.Unix(),
Timeout: int64(timeout.Seconds()),
UniqueKey: uniqueKey,
GroupKey: opt.group,
Retention: int64(opt.retention.Seconds()),
}
now := time.Now()
var state base.TaskState
if opt.processAt.After(now) {
err = c.schedule(ctx, msg, opt.processAt, opt.uniqueTTL)
state = base.TaskStateScheduled
} else if opt.group != "" {
// Use zero value for processAt since we don't know when the task will be aggregated and processed.
opt.processAt = time.Time{}
err = c.addToGroup(ctx, msg, opt.group, opt.uniqueTTL)
state = base.TaskStateAggregating
} else {
opt.processAt = now
err = c.enqueue(ctx, msg, opt.uniqueTTL)
state = base.TaskStatePending
}
switch {
case errors.Is(err, errors.ErrDuplicateTask):
return nil, fmt.Errorf("%w", ErrDuplicateTask)
case errors.Is(err, errors.ErrTaskIdConflict):
return nil, fmt.Errorf("%w", ErrTaskIDConflict)
case err != nil:
return nil, err
}
return newTaskInfo(msg, state, opt.processAt, nil), nil
}
// BatchEnqueueResult holds the result of enqueuing a single task within a batch.
type BatchEnqueueResult struct {
TaskInfo *TaskInfo
Err error
}
// BatchEnqueueContext enqueues multiple tasks in a single Redis pipeline round-trip,
// returning a per-task result slice aligned with the input tasks slice.
//
// # Atomicity Guarantees
//
// There is no all-or-nothing guarantee across the batch. Each task is executed as
// an independent Lua script inside a Redis pipeline. Individual scripts are atomic
// (the existence check, hash write, and list/sorted-set push for one task cannot
// be partially applied), but the pipeline as a whole is not wrapped in a
// MULTI/EXEC transaction. This means:
//
// - Partial success is possible: some tasks may be enqueued while others are not.
// - A task whose ID already exists in Redis is silently skipped (treated as a
// no-op by the Lua script), and its result will still show success.
// - If the Redis pipeline call itself fails (e.g. connection lost, context
// cancelled), every task that passed client-side validation receives that
// error — none of them can be assumed to have been enqueued.
//
// # Validation Errors (pre-pipeline)
//
// The following are caught before any Redis call and rejected in the
// corresponding BatchEnqueueResult.Err without affecting other tasks:
//
// - nil task
// - empty task type name
// - invalid options
// - group tasks (not supported in batch mode)
// - unique tasks (not supported in batch mode)
//
// # Supported Task Types
//
// Immediate and scheduled (via [ProcessAt] or [ProcessIn]) tasks are supported.
// Group and unique tasks are rejected as described above.
func (c *Client) BatchEnqueueContext(ctx context.Context, tasks []*Task, opts ...Option) []BatchEnqueueResult {
results := make([]BatchEnqueueResult, len(tasks))
if len(tasks) == 0 {
return results
}
type itemMeta struct {
state base.TaskState
processAt time.Time
}
items := make([]base.BatchEnqueueItem, 0, len(tasks))
itemIndexes := make([]int, 0, len(tasks))
itemMetas := make([]itemMeta, 0, len(tasks))
for i, task := range tasks {
if task == nil {
results[i] = BatchEnqueueResult{Err: fmt.Errorf("task cannot be nil")}
continue
}
if strings.TrimSpace(task.Type()) == "" {
results[i] = BatchEnqueueResult{Err: fmt.Errorf("task typename cannot be empty")}
continue
}
merged := append(task.opts, opts...)
opt, err := composeOptions(merged...)
if err != nil {
results[i] = BatchEnqueueResult{Err: err}
continue
}
if opt.group != "" {
results[i] = BatchEnqueueResult{Err: fmt.Errorf("batch enqueue does not support group tasks")}
continue
}
if opt.uniqueTTL > 0 {
results[i] = BatchEnqueueResult{Err: fmt.Errorf("batch enqueue does not support unique tasks")}
continue
}
deadline := noDeadline
if !opt.deadline.IsZero() {
deadline = opt.deadline
}
timeout := noTimeout
if opt.timeout != 0 {
timeout = opt.timeout
}
if deadline.Equal(noDeadline) && timeout == noTimeout {
timeout = defaultTimeout
}
msg := &base.TaskMessage{
ID: opt.taskID,
Type: task.Type(),
Payload: task.Payload(),
Headers: task.Headers(),
Queue: opt.queue,
Retry: opt.retry,
Deadline: deadline.Unix(),
Timeout: int64(timeout.Seconds()),
Retention: int64(opt.retention.Seconds()),
}
now := time.Now()
scheduled := opt.processAt.After(now)
item := base.BatchEnqueueItem{Msg: msg}
var meta itemMeta
if scheduled {
item.ProcessAt = opt.processAt
meta = itemMeta{state: base.TaskStateScheduled, processAt: opt.processAt}
} else {
meta = itemMeta{state: base.TaskStatePending, processAt: now}
}
items = append(items, item)
itemIndexes = append(itemIndexes, i)
itemMetas = append(itemMetas, meta)
}
if len(items) == 0 {
return results
}
_, err := c.broker.BatchEnqueue(ctx, items)
if err != nil {
for _, idx := range itemIndexes {
results[idx] = BatchEnqueueResult{Err: err}
}
return results
}
for j, idx := range itemIndexes {
info := newTaskInfo(items[j].Msg, itemMetas[j].state, itemMetas[j].processAt, nil)
results[idx] = BatchEnqueueResult{TaskInfo: info}
}
return results
}
// Ping performs a ping against the redis connection.
func (c *Client) Ping() error {
return c.broker.Ping()
}
func (c *Client) enqueue(ctx context.Context, msg *base.TaskMessage, uniqueTTL time.Duration) error {
if uniqueTTL > 0 {
return c.broker.EnqueueUnique(ctx, msg, uniqueTTL)
}
return c.broker.Enqueue(ctx, msg)
}
func (c *Client) schedule(ctx context.Context, msg *base.TaskMessage, t time.Time, uniqueTTL time.Duration) error {
if uniqueTTL > 0 {
ttl := time.Until(t.Add(uniqueTTL))
return c.broker.ScheduleUnique(ctx, msg, t, ttl)
}
return c.broker.Schedule(ctx, msg, t)
}
func (c *Client) addToGroup(ctx context.Context, msg *base.TaskMessage, group string, uniqueTTL time.Duration) error {
if uniqueTTL > 0 {
return c.broker.AddToGroupUnique(ctx, msg, group, uniqueTTL)
}
return c.broker.AddToGroup(ctx, msg, group)
}