v0.4.0

[ci skip] Update changelog
Use int as priority value type.
2026-05-13 21:16:02 +08:00 · 2020-02-13 06:25:27 -08:00 · 2020-02-13 06:23:51 -08:00 · 2020-02-13 06:23:51 -08:00 · 2020-02-13 06:23:51 -08:00 · 2020-02-13 06:23:51 -08:00
37 changed files with 2264 additions and 860 deletions
--- a/.travis.yml
+++ b/.travis.yml
@@ -7,3 +7,5 @@ env:
 go: [1.12.x, 1.13.x]
 services:
  - redis-server
+after_success:
+  - bash ./.travis/benchcmp.sh
--- a/.travis/benchcmp.sh
+++ b/.travis/benchcmp.sh
@@ -0,0 +1,15 @@
+if [ "${TRAVIS_PULL_REQUEST_BRANCH:-$TRAVIS_BRANCH}" != "master" ]; then
+    REMOTE_URL="$(git config --get remote.origin.url)";
+    cd ${TRAVIS_BUILD_DIR}/.. && \
+    git clone ${REMOTE_URL} "${TRAVIS_REPO_SLUG}-bench" && \
+    cd "${TRAVIS_REPO_SLUG}-bench" && \
+    # Benchmark master
+    git checkout master && \
+    go test -run=XXX -bench=. ./... > master.txt && \
+    # Benchmark feature branch
+    git checkout ${TRAVIS_COMMIT} && \
+    go test -run=XXX -bench=. ./... > feature.txt && \
+    go get -u golang.org/x/tools/cmd/benchcmp && \
+    # compare two benchmarks
+    benchcmp master.txt feature.txt;
+fi
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,35 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

 ## [Unreleased]

+## [0.4.0] - 2020-02-13
+
+### Changed
+
+- `Handler` interface has changed. `ProcessTask` method takes two arguments `context.Context` and `*asynq.Task`
+- `Queues` field in `Config` has change from `map[string]uint` to `map[string]int`
+
+### Added
+
+- `Client` can optionally schedule task with `asynq.Timeout(duration)` to specify timeout duration for task. Default is no timeout.
+- `asynqmon cancel [task id]` will send a cancelation signal to the goroutine processing the speicified task.
+
+## [0.3.0] - 2020-02-04
+
+### Added
+
+- `asynqmon ps` was added to list all background worker processes
+
+## [0.2.2] - 2020-01-26
+
+### Fixed
+
+- Fixed restoring unfinished tasks back to correct queues.
+
+### Changed
+
+- `asynqmon ls` command is now paginated (default 30 tasks from first page)
+- `asynqmon ls enqueued:[queue name]` requires queue name to be specified
+
 ## [0.2.1] - 2020-01-22

 ### Fixed
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -0,0 +1,51 @@
+# Contributing
+
+Thanks for your interest in contributing to Asynq!
+We are open to, and grateful for, any contributions made by the community.
+
+## Reporting Bugs
+
+Have a look at our [issue tracker](https://github.com/hibiken/asynq/issues). If you can't find an issue (open or closed)
+describing your problem (or a very similar one) there, please open a new issue with
+the following details:
+
+- Which versions of Go and Redis are you using?
+- What are you trying to accomplish?
+- What is the full error you are seeing?
+- How can we reproduce this?
+  - Please quote as much of your code as needed to reproduce (best link to a
+    public repository or Gist)
+
+## Getting Help
+
+We run a [Gitter
+channel](https://gitter.im/go-asynq/community) where you can ask questions and
+get help. Feel free to ask there before opening a GitHub issue.
+
+## Submitting Feature Requests
+
+If you can't find an issue (open or closed) describing your idea on our [issue
+tracker](https://github.com/hibiken/asynq/issues), open an issue. Adding answers to the following
+questions in your description is +1:
+
+- What do you want to do, and how do you expect Asynq to support you with that?
+- How might this be added to Asynq?
+- What are possible alternatives?
+- Are there any disadvantages?
+
+Thank you! We'll try to respond as quickly as possible.
+
+## Contributing Code
+
+1. Fork this repo
+2. Download your fork `git clone https://github.com/your-username/asynq && cd asynq`
+3. Create your branch `git checkout -b your-branch-name`
+4. Make and commit your changes
+5. Push the branch `git push origin your-branch-name`
+6. Create a new pull request
+
+Please try to keep your pull request focused in scope and avoid including unrelated commits.
+
+After you have submitted your pull request, we'll try to get back to you as soon as possible. We may suggest some changes or improvements.
+
+Thank you for contributing!
--- a/README.md
+++ b/README.md
@@ -6,39 +6,101 @@
 [![GoDoc](https://godoc.org/github.com/hibiken/asynq?status.svg)](https://godoc.org/github.com/hibiken/asynq)
 [![Gitter chat](https://badges.gitter.im/go-asynq/gitter.svg)](https://gitter.im/go-asynq/community)

-Simple and efficent asynchronous task processing library in Go.
+Asynq is a simple Go library for queueing tasks and processing them in the background with workers.  
+It is backed by Redis and it is designed to have a low barrier to entry. It should be integrated in your web stack easily.

-**Important Note**: Current major version is zero (v0.x.x) to accomodate rapid development and fast iteration while getting early feedback from users. The public API could change without a major version update before the release of verson 1.0.0.
+**Important Note**: Current major version is zero (v0.x.x) to accomodate rapid development and fast iteration while getting early feedback from users. The public API could change without a major version update before v1.0.0 release.

-## Table of Contents
+![Gif](/docs/assets/demo.gif)

- [Overview](#overview)
- [Requirements](#requirements)
- [Installation](#installation)
- [Getting Started](#getting-started)
- [Monitoring CLI](#monitoring-cli)
- [Acknowledgements](#acknowledgements)
- [License](#license)
+## Installation

-## Overview
+To install `asynq` library, run the following command:

-![Gif](/docs/assets/asynqmon_stats.gif)
+```sh
+go get -u github.com/hibiken/asynq
+```

-Asynq provides a simple interface to asynchronous task processing.
+## Quickstart

-It also ships with a tool to monitor the queues and take manual actions if needed.
+First, make sure you are running a Redis server locally.

-Asynq provides:
+```sh
+$ redis-server
+```

- Clear separation of task producer and consumer
- Ability to schedule task processing in the future
- Automatic retry of failed tasks with exponential backoff
- [Automatic failover](https://github.com/hibiken/asynq/wiki/Automatic-Failover) using Redis sentinels
- [Ability to configure](https://github.com/hibiken/asynq/wiki/Task-Retry) max retry count per task
- Ability to configure max number of worker goroutines to process tasks
- Support for [priority queues](https://github.com/hibiken/asynq/wiki/Priority-Queues)
- [Unix signal handling](https://github.com/hibiken/asynq/wiki/Signals) to gracefully shutdown background processing
- [CLI tool](/tools/asynqmon/README.md) to query and mutate queues state for mointoring and administrative purposes
+To create and schedule tasks, use `Client` and provide a task and when to process the task.
+
+```go
+func main() {
+    r := &asynq.RedisClientOpt{
+        Addr: "localhost:6379",
+    }
+
+    client := asynq.NewClient(r)
+
+    // Create a task with task type and payload
+    t1 := asynq.NewTask("send_welcome_email", map[string]interface{}{"user_id": 42})
+
+    t2 := asynq.NewTask("send_reminder_email", map[string]interface{}{"user_id": 42})
+
+    // Process immediately
+    err := client.Schedule(t1, time.Now())
+
+    // Process 24 hrs later
+    err = client.Schedule(t2, time.Now().Add(24 * time.Hour))
+
+    // If processing fails, retry up to 10 times (Default is 25)
+    err = client.Schedule(t1, time.Now(), asynq.Retry(10))
+
+    // Use custom queue called "critical"
+    err = client.Schedule(t1, time.Now(), asynq.Queue("critical"))
+
+    // Use timeout to specify how long a task may run (Default is no limit)
+    err = client.Schedule(t1, time.Now(), asynq.Timeout(30 * time.Second))
+}
+```
+
+To start the background workers, use `Background` and provide your `Handler` to process the tasks.
+
+```go
+func main() {
+    r := &asynq.RedisClientOpt{
+        Addr: "localhost:6379",
+    }
+
+    bg := asynq.NewBackground(r, &asynq.Config{
+        // Specify how many concurrent workers to use
+        Concurrency: 10,
+        // You can optionally create multiple queues with different priority.
+        Queues: map[string]int{
+            "critical": 6,
+            "default":  3,
+            "low":      1,
+        },
+        // See the godoc for other configuration options
+    })
+
+    bg.Run(handler)
+}
+```
+
+`Handler` is an interface with one method `ProcessTask` with the following signature.
+
+```go
+// ProcessTask should return nil if the processing of a task
+// is successful.
+//
+// If ProcessTask return a non-nil error or panics, the task
+// will be retried after delay.
+type Handler interface {
+    ProcessTask(context.Context, *asynq.Task) error
+}
+```
+
+For a more detailed walk-through of the library, see our [Getting Started Guide](https://github.com/hibiken/asynq/wiki/Getting-Started).
+
+To Learn more about `asynq` features and APIs, see our [Wiki pages](https://github.com/hibiken/asynq/wiki) and [godoc](https://godoc.org/github.com/hibiken/asynq).

 ## Requirements

@@ -47,235 +109,22 @@ Asynq provides:
 | [Redis](https://redis.io/) | v2.8+   |
 | [Go](https://golang.org/)  | v1.12+  |

-## Installation
+## Command Line Tool

-To install both `asynq` library and `asynqmon` CLI tool, run the following command:
+Asynq ships with a command line tool to inspect the state of queues and tasks.

-```
-go get -u github.com/hibiken/asynq
+To install, run the following command:
+
+```sh
 go get -u github.com/hibiken/asynq/tools/asynqmon
 ```

-## Getting Started
+For details on how to use the tool, refer to the tool's [README](/tools/asynqmon/README.md).

-In this quick tour of `asynq`, we are going to create two programs.
+## Contributing

- `producer.go` will create and schedule tasks to be processed asynchronously by the consumer.
- `consumer.go` will process the tasks created by the producer.
-
-**This guide assumes that you are running a Redis server at `localhost:6379`**.
-Before we start, make sure you have Redis installed and running.
-
-1. Import `asynq` in both files.
-
-```go
-import "github.com/hibiken/asynq"
-```
-
-2. Asynq uses Redis as a message broker.
-   Use one of `RedisConnOpt` types to specify how to connect to Redis.
-   We are going to use `RedisClientOpt` here.
-
-```go
-// both in producer.go and consumer.go
-var redis = &asynq.RedisClientOpt{
-    Addr: "localhost:6379",
-    // Omit if no password is required
-    Password: "mypassword",
-    // Use a dedicated db number for asynq.
-    // By default, Redis offers 16 databases (0..15)
-    DB: 0,
-}
-```
-
-3. In `producer.go`, create a `Client` instance to create and schedule tasks.
-
-```go
-// producer.go
-func main() {
-    client := asynq.NewClient(redis)
-
-    // Create a task with typename and payload.
-    t1 := asynq.NewTask(
-        "send_welcome_email",
-        map[string]interface{}{"user_id": 42})
-
-    t2 := asynq.NewTask(
-        "send_reminder_email",
-        map[string]interface{}{"user_id": 42})
-
-    // Process the task immediately.
-    err := client.Schedule(t1, time.Now())
-    if err != nil {
-        log.Fatal(err)
-    }
-
-    // Process the task 24 hours later.
-    err = client.Schedule(t2, time.Now().Add(24 * time.Hour))
-    if err != nil {
-        log.Fatal(err)
-    }
-}
-```
-
-4. In `consumer.go`, create a `Background` instance to process tasks.
-
-```go
-// consumer.go
-func main() {
-    bg := asynq.NewBackground(redis, &asynq.Config{
-        Concurrency: 10,
-    })
-
-    bg.Run(handler)
-}
-```
-
-The argument to `(*asynq.Background).Run` is an interface `asynq.Handler` which has one method `ProcessTask`.
-
-```go
-// ProcessTask should return nil if the processing of a task
-// is successful.
-//
-// If ProcessTask return a non-nil error or panics, the task
-// will be retried.
-type Handler interface {
-    ProcessTask(*Task) error
-}
-```
-
-The simplest way to implement a handler is to define a function with the same signature and use `asynq.HandlerFunc` adapter type when passing it to `Run`.
-
-```go
-func handler(t *asynq.Task) error {
-    switch t.Type {
-    case "send_welcome_email":
-        id, err := t.Payload.GetInt("user_id")
-        if err != nil {
-            return err
-        }
-        fmt.Printf("Send Welcome Email to User %d\n", id)
-
-    case "send_reminder_email":
-        id, err := t.Payload.GetInt("user_id")
-        if err != nil {
-            return err
-        }
-        fmt.Printf("Send Reminder Email to User %d\n", id)
-
-    default:
-        return fmt.Errorf("unexpected task type: %s", t.Type)
-    }
-    return nil
-}
-
-func main() {
-    bg := asynq.NewBackground(redis, &asynq.Config{
-        Concurrency: 10,
-    })
-
-    // Use asynq.HandlerFunc adapter for a handler function
-    bg.Run(asynq.HandlerFunc(handler))
-}
-```
-
-We could kep adding cases to this handler function, but in a realistic application, it's convenient to define the logic for each case in a separate function.
-
-To refactor our code, let's create a simple dispatcher which maps task type to its handler.
-
-```go
-// consumer.go
-
-// Dispatcher is used to dispatch tasks to registered handlers.
-type Dispatcher struct {
-    mapping map[string]asynq.HandlerFunc
-}
-
-// HandleFunc registers a task handler
-func (d *Dispatcher) HandleFunc(taskType string, fn asynq.HandlerFunc) {
-    d.mapping[taskType] = fn
-}
-
-// ProcessTask processes a task.
-//
-// NOTE: Dispatcher satisfies asynq.Handler interface.
-func (d *Dispatcher) ProcessTask(task *asynq.Task) error {
-    fn, ok := d.mapping[task.Type]
-    if !ok {
-        return fmt.Errorf("no handler registered for %q", task.Type)
-    }
-    return fn(task)
-}
-
-func main() {
-    d := &Dispatcher{mapping: make(map[string]asynq.HandlerFunc)}
-    d.HandleFunc("send_welcome_email", sendWelcomeEmail)
-    d.HandleFunc("send_reminder_email", sendReminderEmail)
-
-    bg := asynq.NewBackground(redis, &asynq.Config{
-        Concurrency: 10,
-    })
-    bg.Run(d)
-}
-
-func sendWelcomeEmail(t *asynq.Task) error {
-    id, err := t.Payload.GetInt("user_id")
-    if err != nil {
-        return err
-    }
-    fmt.Printf("Send Welcome Email to User %d\n", id)
-    return nil
-}
-
-func sendReminderEmail(t *asynq.Task) error {
-    id, err := t.Payload.GetInt("user_id")
-    if err != nil {
-        return err
-    }
-    fmt.Printf("Send Welcome Email to User %d\n", id)
-    return nil
-}
-```
-
-Now that we have both task producer and consumer, we can run both programs.
-
-```sh
-go run consumer.go
-```
-
-**Note**: This will not exit until you send a signal to terminate the program. See [Signal Wiki page](https://github.com/hibiken/asynq/wiki/Signals) for best practice on how to safely terminate background processing.
-
-With our consumer running, also run
-
-```sh
-go run producer.go
-```
-
-This will create a task and the first task will get processed immediately by the consumer. The second task will be processed 24 hours later.
-
-Let's use `asynqmon` tool to inspect the tasks.
-
-```sh
-asynqmon stats
-```
-
-This command will show the number of tasks in each state and stats for the current date as well as redis information.
-
-To understand the meaning of each state, see [Life of a Task Wiki page](https://github.com/hibiken/asynq/wiki/Life-of-a-Task).
-
-For in-depth guide on `asynqmon` tool, see the [README](/tools/asynqmon/README.md) for the CLI.
-
-This was a quick tour of `asynq` basics. To see all of its features such as **[priority queues](https://github.com/hibiken/asynq/wiki/Priority-Queues)** and **[custom retry](https://github.com/hibiken/asynq/wiki/Task-Retry)**, see [the Wiki page](https://github.com/hibiken/asynq/wiki).
-
-## Monitoring CLI
-
-Asynq ships with a CLI tool to inspect the state of queues and tasks.
-
-To install the CLI, run the following command:
-
-    go get github.com/hibiken/asynq/tools/asynqmon
-
-For details on how to use the tool, see the [README](/tools/asynqmon/README.md) for the asynqmon CLI.
+We are open to, and grateful for, any contributions (Github issues/pull-requests, feedback on Gitter channel, etc) made by the community.
+Please see the [Contribution Guide](/CONTRIBUTING.md) before contributing.

 ## Acknowledgements

--- a/asynq.go
+++ b/asynq.go
@@ -11,7 +11,7 @@ import (
 	"github.com/go-redis/redis/v7"
 )

-// Task represents a task to be performed.
+// Task represents a unit of work to be performed.
 type Task struct {
 	// Type indicates the type of task to be performed.
 	Type string
@@ -20,10 +20,9 @@ type Task struct {
 	Payload Payload
 }

-// NewTask returns a new Task. The typename and payload argument set Type
-// and Payload field respectively.
+// NewTask returns a new Task given a type name and payload data.
 //
-// The payload must be serializable to JSON.
+// The payload values must be serializable.
 func NewTask(typename string, payload map[string]interface{}) *Task {
 	return &Task{
 		Type:    typename,
@@ -31,13 +30,11 @@ func NewTask(typename string, payload map[string]interface{}) *Task {
 	}
 }

-// RedisConnOpt is a discriminated union of redis-client-option types.
+// RedisConnOpt is a discriminated union of types that represent Redis connection configuration option.
 //
 // RedisConnOpt represents a sum of following types:
 //
 // RedisClientOpt | *RedisClientOpt | RedisFailoverClientOpt | *RedisFailoverClientOpt
-//
-// Passing unexpected type to a RedisConnOpt variable can cause panic.
 type RedisConnOpt interface{}

 // RedisClientOpt is used to create a redis client that connects
@@ -53,7 +50,7 @@ type RedisClientOpt struct {
 	// Redis server password.
 	Password string

-	// Redis DB to select after connecting to the server.
+	// Redis DB to select after connecting to a server.
 	// See: https://redis.io/commands/select.
 	DB int

@@ -61,13 +58,13 @@ type RedisClientOpt struct {
 	// Default is 10 connections per every CPU as reported by runtime.NumCPU.
 	PoolSize int

-	// TLS Config used to connect to the server.
+	// TLS Config used to connect to a server.
 	// TLS will be negotiated only if this field is set.
 	TLSConfig *tls.Config
 }

 // RedisFailoverClientOpt is used to creates a redis client that talks
-// to redis sentinels for service discovery and has automatic failover
+// to redis sentinels for service discovery and has an automatic failover
 // capability.
 type RedisFailoverClientOpt struct {
 	// Redis master name that monitored by sentinels.
@@ -84,7 +81,7 @@ type RedisFailoverClientOpt struct {
 	// Redis server password.
 	Password string

-	// Redis DB to select after connecting to the server.
+	// Redis DB to select after connecting to a server.
 	// See: https://redis.io/commands/select.
 	DB int

@@ -92,11 +89,14 @@ type RedisFailoverClientOpt struct {
 	// Default is 10 connections per every CPU as reported by runtime.NumCPU.
 	PoolSize int

-	// TLS Config used to connect to the server.
+	// TLS Config used to connect to a server.
 	// TLS will be negotiated only if this field is set.
 	TLSConfig *tls.Config
 }

+// createRedisClient returns a redis client given a redis connection configuration.
+//
+// Passing an unexpected type as a RedisConnOpt argument will cause panic.
 func createRedisClient(r RedisConnOpt) *redis.Client {
 	switch r := r.(type) {
 	case *RedisClientOpt:
--- a/background.go
+++ b/background.go
@@ -5,6 +5,7 @@
 package asynq

 import (
+	"context"
 	"fmt"
 	"math"
 	"math/rand"
@@ -33,17 +34,20 @@ type Background struct {
 	mu      sync.Mutex
 	running bool

-	rdb       *rdb.RDB
-	scheduler *scheduler
-	processor *processor
-	syncer    *syncer
+	pinfo       *base.ProcessInfo
+	rdb         *rdb.RDB
+	scheduler   *scheduler
+	processor   *processor
+	syncer      *syncer
+	heartbeater *heartbeater
+	subscriber  *subscriber
 }

 // Config specifies the background-task processing behavior.
 type Config struct {
 	// Maximum number of concurrent processing of tasks.
 	//
-	// If set to zero or negative value, NewBackground will overwrite the value to one.
+	// If set to a zero or negative value, NewBackground will overwrite the value to one.
 	Concurrency int

 	// Function to calculate retry delay for a failed task.
@@ -55,15 +59,15 @@ type Config struct {
 	// t is the task in question.
 	RetryDelayFunc func(n int, e error, t *Task) time.Duration

-	// List of queues to process with given priority level. Keys are the names of the
-	// queues and values are associated priority level.
+	// List of queues to process with given priority value. Keys are the names of the
+	// queues and values are associated priority value.
 	//
 	// If set to nil or not specified, the background will process only the "default" queue.
 	//
 	// Priority is treated as follows to avoid starving low priority queues.
 	//
 	// Example:
-	// Queues: map[string]uint{
+	// Queues: map[string]int{
 	//     "critical": 6,
 	//     "default":  3,
 	//     "low":      1,
@@ -71,7 +75,9 @@ type Config struct {
 	// With the above config and given that all queues are not empty, the tasks
 	// in "critical", "default", "low" should be processed 60%, 30%, 10% of
 	// the time respectively.
-	Queues map[string]uint
+	//
+	// If a queue has a zero or negative priority value, the queue will be ignored.
+	Queues map[string]int

 	// StrictPriority indicates whether the queue priority should be treated strictly.
 	//
@@ -88,7 +94,7 @@ func defaultDelayFunc(n int, e error, t *Task) time.Duration {
 	return time.Duration(s) * time.Second
 }

-var defaultQueueConfig = map[string]uint{
+var defaultQueueConfig = map[string]int{
 	base.DefaultQueueName: 1,
 }

@@ -103,22 +109,39 @@ func NewBackground(r RedisConnOpt, cfg *Config) *Background {
 	if delayFunc == nil {
 		delayFunc = defaultDelayFunc
 	}
-	queues := cfg.Queues
-	if queues == nil || len(queues) == 0 {
+	queues := make(map[string]int)
+	for qname, p := range cfg.Queues {
+		if p > 0 {
+			queues[qname] = p
+		}
+	}
+	if len(queues) == 0 {
 		queues = defaultQueueConfig
 	}
-	qcfg := normalizeQueueCfg(queues)

-	syncRequestCh := make(chan *syncRequest)
-	syncer := newSyncer(syncRequestCh, 5*time.Second)
+	host, err := os.Hostname()
+	if err != nil {
+		host = "unknown-host"
+	}
+	pid := os.Getpid()
+
+	pinfo := base.NewProcessInfo(host, pid, n, queues, cfg.StrictPriority)
 	rdb := rdb.NewRDB(createRedisClient(r))
-	scheduler := newScheduler(rdb, 5*time.Second, qcfg)
-	processor := newProcessor(rdb, n, qcfg, cfg.StrictPriority, delayFunc, syncRequestCh)
+	syncRequestCh := make(chan *syncRequest)
+	cancelations := base.NewCancelations()
+	syncer := newSyncer(syncRequestCh, 5*time.Second)
+	heartbeater := newHeartbeater(rdb, pinfo, 5*time.Second)
+	scheduler := newScheduler(rdb, 5*time.Second, queues)
+	processor := newProcessor(rdb, pinfo, delayFunc, syncRequestCh, cancelations)
+	subscriber := newSubscriber(rdb, cancelations)
 	return &Background{
-		rdb:       rdb,
-		scheduler: scheduler,
-		processor: processor,
-		syncer:    syncer,
+		pinfo:       pinfo,
+		rdb:         rdb,
+		scheduler:   scheduler,
+		processor:   processor,
+		syncer:      syncer,
+		heartbeater: heartbeater,
+		subscriber:  subscriber,
 	}
 }

@@ -130,18 +153,18 @@ func NewBackground(r RedisConnOpt, cfg *Config) *Background {
 // If ProcessTask return a non-nil error or panics, the task
 // will be retried after delay.
 type Handler interface {
-	ProcessTask(*Task) error
+	ProcessTask(context.Context, *Task) error
 }

 // The HandlerFunc type is an adapter to allow the use of
 // ordinary functions as a Handler. If f is a function
 // with the appropriate signature, HandlerFunc(f) is a
 // Handler that calls f.
-type HandlerFunc func(*Task) error
+type HandlerFunc func(context.Context, *Task) error

-// ProcessTask calls fn(task)
-func (fn HandlerFunc) ProcessTask(task *Task) error {
-	return fn(task)
+// ProcessTask calls fn(ctx, task)
+func (fn HandlerFunc) ProcessTask(ctx context.Context, task *Task) error {
+	return fn(ctx, task)
 }

 // Run starts the background-task processing and blocks until
@@ -165,6 +188,7 @@ func (bg *Background) Run(handler Handler) {
 		sig := <-sigs
 		if sig == syscall.SIGTSTP {
 			bg.processor.stop()
+			bg.pinfo.SetState("stopped")
 			continue
 		}
 		break
@@ -184,6 +208,8 @@ func (bg *Background) start(handler Handler) {
 	bg.running = true
 	bg.processor.handler = handler

+	bg.heartbeater.start()
+	bg.subscriber.start()
 	bg.syncer.start()
 	bg.scheduler.start()
 	bg.processor.start()
@@ -202,42 +228,13 @@ func (bg *Background) stop() {
 	// Note: processor and all worker goroutines need to be exited
 	// before shutting down syncer to avoid goroutine leak.
 	bg.syncer.terminate()
+	bg.subscriber.terminate()
+	bg.heartbeater.terminate()

+	bg.rdb.ClearProcessInfo(bg.pinfo)
 	bg.rdb.Close()
 	bg.processor.handler = nil
 	bg.running = false

 	logger.info("Bye!")
 }
-
-// normalizeQueueCfg divides priority numbers by their
-// greatest common divisor.
-func normalizeQueueCfg(queueCfg map[string]uint) map[string]uint {
-	var xs []uint
-	for _, x := range queueCfg {
-		xs = append(xs, x)
-	}
-	d := gcd(xs...)
-	res := make(map[string]uint)
-	for q, x := range queueCfg {
-		res[q] = x / d
-	}
-	return res
-}
-
-func gcd(xs ...uint) uint {
-	fn := func(x, y uint) uint {
-		for y > 0 {
-			x, y = y, x%y
-		}
-		return x
-	}
-	res := xs[0]
-	for i := 0; i < len(xs); i++ {
-		res = fn(xs[i], res)
-		if res == 1 {
-			return 1
-		}
-	}
-	return res
-}
--- a/background_test.go
+++ b/background_test.go
@@ -5,6 +5,7 @@
 package asynq

 import (
+	"context"
 	"testing"
 	"time"

@@ -27,7 +28,7 @@ func TestBackground(t *testing.T) {
 	})

 	// no-op handler
-	h := func(task *Task) error {
+	h := func(ctx context.Context, task *Task) error {
 		return nil
 	}

@@ -42,16 +43,16 @@ func TestBackground(t *testing.T) {

 func TestGCD(t *testing.T) {
 	tests := []struct {
-		input []uint
-		want  uint
+		input []int
+		want  int
 	}{
-		{[]uint{6, 2, 12}, 2},
-		{[]uint{3, 3, 3}, 3},
-		{[]uint{6, 3, 1}, 1},
-		{[]uint{1}, 1},
-		{[]uint{1, 0, 2}, 1},
-		{[]uint{8, 0, 4}, 4},
-		{[]uint{9, 12, 18, 30}, 3},
+		{[]int{6, 2, 12}, 2},
+		{[]int{3, 3, 3}, 3},
+		{[]int{6, 3, 1}, 1},
+		{[]int{1}, 1},
+		{[]int{1, 0, 2}, 1},
+		{[]int{8, 0, 4}, 4},
+		{[]int{9, 12, 18, 30}, 3},
 	}

 	for _, tc := range tests {
@@ -64,46 +65,46 @@ func TestGCD(t *testing.T) {

 func TestNormalizeQueueCfg(t *testing.T) {
 	tests := []struct {
-		input map[string]uint
-		want  map[string]uint
+		input map[string]int
+		want  map[string]int
 	}{
 		{
-			input: map[string]uint{
+			input: map[string]int{
 				"high":    100,
 				"default": 20,
 				"low":     5,
 			},
-			want: map[string]uint{
+			want: map[string]int{
 				"high":    20,
 				"default": 4,
 				"low":     1,
 			},
 		},
 		{
-			input: map[string]uint{
+			input: map[string]int{
 				"default": 10,
 			},
-			want: map[string]uint{
+			want: map[string]int{
 				"default": 1,
 			},
 		},
 		{
-			input: map[string]uint{
+			input: map[string]int{
 				"critical": 5,
 				"default":  1,
 			},
-			want: map[string]uint{
+			want: map[string]int{
 				"critical": 5,
 				"default":  1,
 			},
 		},
 		{
-			input: map[string]uint{
+			input: map[string]int{
 				"critical": 6,
 				"default":  3,
 				"low":      0,
 			},
-			want: map[string]uint{
+			want: map[string]int{
 				"critical": 2,
 				"default":  1,
 				"low":      0,
--- a/benchmark_test.go
+++ b/benchmark_test.go
@@ -5,6 +5,7 @@
 package asynq

 import (
+	"context"
 	"fmt"
 	"math/rand"
 	"sync"
@@ -12,8 +13,7 @@ import (
 	"time"
 )

-// Simple E2E Benchmark testing with no scheduled tasks and
-// no retries.
+// Simple E2E Benchmark testing with no scheduled tasks and retries.
 func BenchmarkEndToEndSimple(b *testing.B) {
 	const count = 100000
 	for n := 0; n < b.N; n++ {
@@ -38,7 +38,7 @@ func BenchmarkEndToEndSimple(b *testing.B) {

 		var wg sync.WaitGroup
 		wg.Add(count)
-		handler := func(t *Task) error {
+		handler := func(ctx context.Context, t *Task) error {
 			wg.Done()
 			return nil
 		}
@@ -83,7 +83,7 @@ func BenchmarkEndToEnd(b *testing.B) {

 		var wg sync.WaitGroup
 		wg.Add(count * 2)
-		handler := func(t *Task) error {
+		handler := func(ctx context.Context, t *Task) error {
 			// randomly fail 1% of tasks
 			if rand.Intn(100) == 1 {
 				return fmt.Errorf(":(")
@@ -101,3 +101,58 @@ func BenchmarkEndToEnd(b *testing.B) {
 		b.StartTimer() // end teardown
 	}
 }
+
+// Simple E2E Benchmark testing with no scheduled tasks and retries with multiple queues.
+func BenchmarkEndToEndMultipleQueues(b *testing.B) {
+	// number of tasks to create for each queue
+	const (
+		highCount    = 20000
+		defaultCount = 20000
+		lowCount     = 20000
+	)
+	for n := 0; n < b.N; n++ {
+		b.StopTimer() // begin setup
+		setup(b)
+		redis := &RedisClientOpt{
+			Addr: redisAddr,
+			DB:   redisDB,
+		}
+		client := NewClient(redis)
+		bg := NewBackground(redis, &Config{
+			Concurrency: 10,
+			Queues: map[string]int{
+				"high":    6,
+				"default": 3,
+				"low":     1,
+			},
+		})
+		// Create a bunch of tasks
+		for i := 0; i < highCount; i++ {
+			t := NewTask(fmt.Sprintf("task%d", i), map[string]interface{}{"data": i})
+			client.Schedule(t, time.Now(), Queue("high"))
+		}
+		for i := 0; i < defaultCount; i++ {
+			t := NewTask(fmt.Sprintf("task%d", i), map[string]interface{}{"data": i})
+			client.Schedule(t, time.Now())
+		}
+		for i := 0; i < lowCount; i++ {
+			t := NewTask(fmt.Sprintf("task%d", i), map[string]interface{}{"data": i})
+			client.Schedule(t, time.Now(), Queue("low"))
+		}
+
+		var wg sync.WaitGroup
+		wg.Add(highCount + defaultCount + lowCount)
+		handler := func(ctx context.Context, t *Task) error {
+			wg.Done()
+			return nil
+		}
+		b.StartTimer() // end setup
+
+		bg.start(HandlerFunc(handler))
+		wg.Wait()
+
+		b.StopTimer() // begin teardown
+		bg.stop()
+		b.StartTimer() // end teardown
+	}
+}
--- a/client.go
+++ b/client.go
@@ -34,8 +34,9 @@ type Option interface{}

 // Internal option representations.
 type (
-	retryOption int
-	queueOption string
+	retryOption   int
+	queueOption   string
+	timeoutOption time.Duration
 )

 // MaxRetry returns an option to specify the max number of times
@@ -56,15 +57,24 @@ func Queue(name string) Option {
 	return queueOption(strings.ToLower(name))
 }

+// Timeout returns an option to specify how long a task may run.
+//
+// Zero duration means no limit.
+func Timeout(d time.Duration) Option {
+	return timeoutOption(d)
+}
+
 type option struct {
-	retry int
-	queue string
+	retry   int
+	queue   string
+	timeout time.Duration
 }

 func composeOptions(opts ...Option) option {
 	res := option{
-		retry: defaultMaxRetry,
-		queue: base.DefaultQueueName,
+		retry:   defaultMaxRetry,
+		queue:   base.DefaultQueueName,
+		timeout: 0,
 	}
 	for _, opt := range opts {
 		switch opt := opt.(type) {
@@ -72,6 +82,8 @@ func composeOptions(opts ...Option) option {
 			res.retry = int(opt)
 		case queueOption:
 			res.queue = string(opt)
+		case timeoutOption:
+			res.timeout = time.Duration(opt)
 		default:
 			// ignore unexpected option
 		}
@@ -99,6 +111,7 @@ func (c *Client) Schedule(task *Task, processAt time.Time, opts ...Option) error
 		Payload: task.Payload.data,
 		Queue:   opt.queue,
 		Retry:   opt.retry,
+		Timeout: opt.timeout.String(),
 	}
 	return c.enqueue(msg, processAt)
 }
--- a/client_test.go
+++ b/client_test.go
@@ -42,6 +42,7 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   defaultMaxRetry,
 						Queue:   "default",
+						Timeout: time.Duration(0).String(),
 					},
 				},
 			},
@@ -60,6 +61,7 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   defaultMaxRetry,
 						Queue:   "default",
+						Timeout: time.Duration(0).String(),
 					},
 					Score: float64(time.Now().Add(2 * time.Hour).Unix()),
 				},
@@ -79,6 +81,7 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   3,
 						Queue:   "default",
+						Timeout: time.Duration(0).String(),
 					},
 				},
 			},
@@ -98,6 +101,7 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   0, // Retry count should be set to zero
 						Queue:   "default",
+						Timeout: time.Duration(0).String(),
 					},
 				},
 			},
@@ -118,6 +122,7 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   10, // Last option takes precedence
 						Queue:   "default",
+						Timeout: time.Duration(0).String(),
 					},
 				},
 			},
@@ -137,6 +142,7 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   defaultMaxRetry,
 						Queue:   "custom",
+						Timeout: time.Duration(0).String(),
 					},
 				},
 			},
@@ -156,6 +162,27 @@ func TestClient(t *testing.T) {
 						Payload: task.Payload.data,
 						Retry:   defaultMaxRetry,
 						Queue:   "high",
+						Timeout: time.Duration(0).String(),
+					},
+				},
+			},
+			wantScheduled: nil, // db is flushed in setup so zset does not exist hence nil
+		},
+		{
+			desc:      "Timeout option sets the timeout duration",
+			task:      task,
+			processAt: time.Now(),
+			opts: []Option{
+				Timeout(20 * time.Second),
+			},
+			wantEnqueued: map[string][]*base.TaskMessage{
+				"default": []*base.TaskMessage{
+					&base.TaskMessage{
+						Type:    task.Type,
+						Payload: task.Payload.data,
+						Retry:   defaultMaxRetry,
+						Queue:   "default",
+						Timeout: (20 * time.Second).String(),
 					},
 				},
 			},
--- a/doc.go
+++ b/doc.go
@@ -3,7 +3,7 @@
 // that can be found in the LICENSE file.

 /*
-Package asynq provides a framework for background task processing.
+Package asynq provides a framework for asynchronous task processing.

 Asynq uses Redis as a message broker. To connect to redis server,
 specify the options using one of RedisConnOpt types.
@@ -45,7 +45,7 @@ Example of a type that implements the Handler interface.
        // ...
    }

-    func (h *TaskHandler) ProcessTask(task *asynq.Task) error {
+    func (h *TaskHandler) ProcessTask(ctx context.Context, task *asynq.Task) error {
        switch task.Type {
        case "send_email":
            id, err := task.Payload.GetInt("user_id")
--- a/docs/assets/asynqmon_history.gif
+++ b/docs/assets/asynqmon_history.gif
--- a/docs/assets/asynqmon_ps.gif
+++ b/docs/assets/asynqmon_ps.gif
--- a/docs/assets/demo.gif
+++ b/docs/assets/demo.gif
--- a/heartbeat.go
+++ b/heartbeat.go
@@ -0,0 +1,67 @@
+// 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 (
+	"time"
+
+	"github.com/hibiken/asynq/internal/base"
+	"github.com/hibiken/asynq/internal/rdb"
+)
+
+// heartbeater is responsible for writing process info to redis periodically to
+// indicate that the background worker process is up.
+type heartbeater struct {
+	rdb *rdb.RDB
+
+	pinfo *base.ProcessInfo
+
+	// channel to communicate back to the long running "heartbeater" goroutine.
+	done chan struct{}
+
+	// interval between heartbeats.
+	interval time.Duration
+}
+
+func newHeartbeater(rdb *rdb.RDB, pinfo *base.ProcessInfo, interval time.Duration) *heartbeater {
+	return &heartbeater{
+		rdb:      rdb,
+		pinfo:    pinfo,
+		done:     make(chan struct{}),
+		interval: interval,
+	}
+}
+
+func (h *heartbeater) terminate() {
+	logger.info("Heartbeater shutting down...")
+	// Signal the heartbeater goroutine to stop.
+	h.done <- struct{}{}
+}
+
+func (h *heartbeater) start() {
+	h.pinfo.SetStarted(time.Now())
+	h.pinfo.SetState("running")
+	go func() {
+		h.beat()
+		for {
+			select {
+			case <-h.done:
+				logger.info("Heartbeater done")
+				return
+			case <-time.After(h.interval):
+				h.beat()
+			}
+		}
+	}()
+}
+
+func (h *heartbeater) beat() {
+	// Note: Set TTL to be long enough so that it won't expire before we write again
+	// and short enough to expire quickly once the process is shut down or killed.
+	err := h.rdb.WriteProcessInfo(h.pinfo, h.interval*2)
+	if err != nil {
+		logger.error("could not write heartbeat data: %v", err)
+	}
+}
--- a/heartbeat_test.go
+++ b/heartbeat_test.go
@@ -0,0 +1,88 @@
+// 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 (
+	"testing"
+	"time"
+
+	"github.com/google/go-cmp/cmp"
+	"github.com/google/go-cmp/cmp/cmpopts"
+	h "github.com/hibiken/asynq/internal/asynqtest"
+	"github.com/hibiken/asynq/internal/base"
+	"github.com/hibiken/asynq/internal/rdb"
+)
+
+func TestHeartbeater(t *testing.T) {
+	r := setup(t)
+	rdbClient := rdb.NewRDB(r)
+
+	tests := []struct {
+		interval    time.Duration
+		host        string
+		pid         int
+		queues      map[string]int
+		concurrency int
+	}{
+		{time.Second, "some.address.ec2.aws.com", 45678, map[string]int{"default": 1}, 10},
+	}
+
+	timeCmpOpt := cmpopts.EquateApproxTime(10 * time.Millisecond)
+	ignoreOpt := cmpopts.IgnoreUnexported(base.ProcessInfo{})
+	for _, tc := range tests {
+		h.FlushDB(t, r)
+
+		pi := base.NewProcessInfo(tc.host, tc.pid, tc.concurrency, tc.queues, false)
+		hb := newHeartbeater(rdbClient, pi, tc.interval)
+
+		want := &base.ProcessInfo{
+			Host:        tc.host,
+			PID:         tc.pid,
+			Queues:      tc.queues,
+			Concurrency: tc.concurrency,
+			Started:     time.Now(),
+			State:       "running",
+		}
+		hb.start()
+
+		// allow for heartbeater to write to redis
+		time.Sleep(tc.interval * 2)
+
+		got, err := rdbClient.ReadProcessInfo(tc.host, tc.pid)
+		if err != nil {
+			t.Errorf("could not read process status from redis: %v", err)
+			hb.terminate()
+			continue
+		}
+
+		if diff := cmp.Diff(want, got, timeCmpOpt, ignoreOpt); diff != "" {
+			t.Errorf("redis stored process status %+v, want %+v; (-want, +got)\n%s", got, want, diff)
+			hb.terminate()
+			continue
+		}
+
+		// state change
+		pi.SetState("stopped")
+
+		// allow for heartbeater to write to redis
+		time.Sleep(tc.interval * 2)
+
+		want.State = "stopped"
+		got, err = rdbClient.ReadProcessInfo(tc.host, tc.pid)
+		if err != nil {
+			t.Errorf("could not read process status from redis: %v", err)
+			hb.terminate()
+			continue
+		}
+
+		if diff := cmp.Diff(want, got, timeCmpOpt, ignoreOpt); diff != "" {
+			t.Errorf("redis stored process status %+v, want %+v; (-want, +got)\n%s", got, want, diff)
+			hb.terminate()
+			continue
+		}
+
+		hb.terminate()
+	}
+}
--- a/internal/asynqtest/asynqtest.go
+++ b/internal/asynqtest/asynqtest.go
@@ -41,6 +41,18 @@ var SortZSetEntryOpt = cmp.Transformer("SortZSetEntries", func(in []ZSetEntry) [
 	return out
 })

+// SortProcessInfoOpt is a cmp.Option to sort base.ProcessInfo for comparing slice of process info.
+var SortProcessInfoOpt = cmp.Transformer("SortProcessInfo", func(in []*base.ProcessInfo) []*base.ProcessInfo {
+	out := append([]*base.ProcessInfo(nil), in...) // Copy input to avoid mutating it
+	sort.Slice(out, func(i, j int) bool {
+		if out[i].Host != out[j].Host {
+			return out[i].Host < out[j].Host
+		}
+		return out[i].PID < out[j].PID
+	})
+	return out
+})
+
 // IgnoreIDOpt is an cmp.Option to ignore ID field in task messages when comparing.
 var IgnoreIDOpt = cmpopts.IgnoreFields(base.TaskMessage{}, "ID")

--- a/internal/base/base.go
+++ b/internal/base/base.go
@@ -6,7 +6,10 @@
 package base

 import (
+	"context"
+	"fmt"
 	"strings"
+	"sync"
 	"time"

 	"github.com/rs/xid"
@@ -17,6 +20,8 @@ const DefaultQueueName = "default"

 // Redis keys
 const (
+	psPrefix        = "asynq:ps:"                    // HASH
+	AllProcesses    = "asynq:ps"                     // ZSET
 	processedPrefix = "asynq:processed:"             // STRING - asynq:processed:<yyyy-mm-dd>
 	failurePrefix   = "asynq:failure:"               // STRING - asynq:failure:<yyyy-mm-dd>
 	QueuePrefix     = "asynq:queues:"                // LIST   - asynq:queues:<qname>
@@ -26,6 +31,7 @@ const (
 	RetryQueue      = "asynq:retry"                  // ZSET
 	DeadQueue       = "asynq:dead"                   // ZSET
 	InProgressQueue = "asynq:in_progress"            // LIST
+	CancelChannel   = "asynq:cancel"                 // PubSub channel
 )

 // QueueKey returns a redis key string for the given queue name.
@@ -45,6 +51,11 @@ func FailureKey(t time.Time) string {
 	return failurePrefix + t.UTC().Format("2006-01-02")
 }

+// ProcessInfoKey returns a redis key string for process info.
+func ProcessInfoKey(hostname string, pid int) string {
+	return fmt.Sprintf("%s%s:%d", psPrefix, hostname, pid)
+}
+
 // TaskMessage is the internal representation of a task with additional metadata fields.
 // Serialized data of this type gets written to redis.
 type TaskMessage struct {
@@ -68,4 +79,102 @@ type TaskMessage struct {

 	// ErrorMsg holds the error message from the last failure.
 	ErrorMsg string
+
+	// Timeout specifies how long a task may run.
+	// The string value should be compatible with time.Duration.ParseDuration.
+	//
+	// Zero means no limit.
+	Timeout string
+}
+
+// ProcessInfo holds information about running background worker process.
+type ProcessInfo struct {
+	mu                sync.Mutex
+	Concurrency       int
+	Queues            map[string]int
+	StrictPriority    bool
+	PID               int
+	Host              string
+	State             string
+	Started           time.Time
+	ActiveWorkerCount int
+}
+
+// NewProcessInfo returns a new instance of ProcessInfo.
+func NewProcessInfo(host string, pid, concurrency int, queues map[string]int, strict bool) *ProcessInfo {
+	return &ProcessInfo{
+		Host:           host,
+		PID:            pid,
+		Concurrency:    concurrency,
+		Queues:         queues,
+		StrictPriority: strict,
+	}
+}
+
+// SetState set the state field of the process info.
+func (p *ProcessInfo) SetState(state string) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	p.State = state
+}
+
+// SetStarted set the started field of the process info.
+func (p *ProcessInfo) SetStarted(t time.Time) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	p.Started = t
+}
+
+// IncrActiveWorkerCount increments active worker count by delta.
+func (p *ProcessInfo) IncrActiveWorkerCount(delta int) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	p.ActiveWorkerCount += delta
+}
+
+// Cancelations is a collection that holds cancel functions for all in-progress tasks.
+//
+// Its methods are safe to be used in multiple goroutines.
+type Cancelations struct {
+	mu          sync.Mutex
+	cancelFuncs map[string]context.CancelFunc
+}
+
+// NewCancelations returns a Cancelations instance.
+func NewCancelations() *Cancelations {
+	return &Cancelations{
+		cancelFuncs: make(map[string]context.CancelFunc),
+	}
+}
+
+// Add adds a new cancel func to the collection.
+func (c *Cancelations) Add(id string, fn context.CancelFunc) {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+	c.cancelFuncs[id] = fn
+}
+
+// Delete deletes a cancel func from the collection given an id.
+func (c *Cancelations) Delete(id string) {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+	delete(c.cancelFuncs, id)
+}
+
+// Get returns a cancel func given an id.
+func (c *Cancelations) Get(id string) context.CancelFunc {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+	return c.cancelFuncs[id]
+}
+
+// GetAll returns all cancel funcs.
+func (c *Cancelations) GetAll() []context.CancelFunc {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+	var res []context.CancelFunc
+	for _, fn := range c.cancelFuncs {
+		res = append(res, fn)
+	}
+	return res
 }
--- a/internal/base/base_test.go
+++ b/internal/base/base_test.go
@@ -5,6 +5,7 @@
 package base

 import (
+	"sync"
 	"testing"
 	"time"
 )
@@ -60,3 +61,48 @@ func TestFailureKey(t *testing.T) {
 		}
 	}
 }
+
+func TestProcessInfoKey(t *testing.T) {
+	tests := []struct {
+		hostname string
+		pid      int
+		want     string
+	}{
+		{"localhost", 9876, "asynq:ps:localhost:9876"},
+		{"127.0.0.1", 1234, "asynq:ps:127.0.0.1:1234"},
+	}
+
+	for _, tc := range tests {
+		got := ProcessInfoKey(tc.hostname, tc.pid)
+		if got != tc.want {
+			t.Errorf("ProcessInfoKey(%s, %d) = %s, want %s", tc.hostname, tc.pid, got, tc.want)
+		}
+	}
+}
+
+// Note: Run this test with -race flag to check for data race.
+func TestProcessInfoSetter(t *testing.T) {
+	pi := NewProcessInfo("localhost", 1234, 8, map[string]int{"default": 1}, false)
+
+	var wg sync.WaitGroup
+
+	wg.Add(3)
+
+	go func() {
+		pi.SetState("runnning")
+		wg.Done()
+	}()
+
+	go func() {
+		pi.SetStarted(time.Now())
+		pi.IncrActiveWorkerCount(1)
+		wg.Done()
+	}()
+
+	go func() {
+		pi.SetState("stopped")
+		wg.Done()
+	}()
+
+	wg.Wait()
+}
--- a/internal/rdb/inspect.go
+++ b/internal/rdb/inspect.go
@@ -86,49 +86,48 @@ type DeadTask struct {
 	Queue        string
 }

+// KEYS[1] -> asynq:queues
+// KEYS[2] -> asynq:in_progress
+// KEYS[3] -> asynq:scheduled
+// KEYS[4] -> asynq:retry
+// KEYS[5] -> asynq:dead
+// KEYS[6] -> asynq:processed:<yyyy-mm-dd>
+// KEYS[7] -> asynq:failure:<yyyy-mm-dd>
+var currentStatsCmd = redis.NewScript(`
+local res = {}
+local queues = redis.call("SMEMBERS", KEYS[1])
+for _, qkey in ipairs(queues) do
+	table.insert(res, qkey)
+	table.insert(res, redis.call("LLEN", qkey))
+end
+table.insert(res, KEYS[2])
+table.insert(res, redis.call("LLEN", KEYS[2]))
+table.insert(res, KEYS[3])
+table.insert(res, redis.call("ZCARD", KEYS[3]))
+table.insert(res, KEYS[4])
+table.insert(res, redis.call("ZCARD", KEYS[4]))
+table.insert(res, KEYS[5])
+table.insert(res, redis.call("ZCARD", KEYS[5]))
+local pcount = 0
+local p = redis.call("GET", KEYS[6])
+if p then
+	pcount = tonumber(p) 
+end
+table.insert(res, "processed")
+table.insert(res, pcount)
+local fcount = 0
+local f = redis.call("GET", KEYS[7])
+if f then
+	fcount = tonumber(f)
+end
+table.insert(res, "failed")
+table.insert(res, fcount)
+return res`)
+
 // CurrentStats returns a current state of the queues.
 func (r *RDB) CurrentStats() (*Stats, error) {
-	// KEYS[1] -> asynq:queues
-	// KEYS[2] -> asynq:in_progress
-	// KEYS[3] -> asynq:scheduled
-	// KEYS[4] -> asynq:retry
-	// KEYS[5] -> asynq:dead
-	// KEYS[6] -> asynq:processed:<yyyy-mm-dd>
-	// KEYS[7] -> asynq:failure:<yyyy-mm-dd>
-	script := redis.NewScript(`
-	local res = {}
-	local queues = redis.call("SMEMBERS", KEYS[1])
-	for _, qkey in ipairs(queues) do
-	  table.insert(res, qkey)
-	  table.insert(res, redis.call("LLEN", qkey))
-	end
-	table.insert(res, KEYS[2])
-	table.insert(res, redis.call("LLEN", KEYS[2]))
-	table.insert(res, KEYS[3])
-	table.insert(res, redis.call("ZCARD", KEYS[3]))
-	table.insert(res, KEYS[4])
-	table.insert(res, redis.call("ZCARD", KEYS[4]))
-	table.insert(res, KEYS[5])
-	table.insert(res, redis.call("ZCARD", KEYS[5]))
-	local pcount = 0
-	local p = redis.call("GET", KEYS[6])
-	if p then
-		pcount = tonumber(p) 
-	end
-	table.insert(res, "processed")
-	table.insert(res, pcount)
-	local fcount = 0
-	local f = redis.call("GET", KEYS[7])
-	if f then
-		fcount = tonumber(f)
-	end
-	table.insert(res, "failed")
-	table.insert(res, fcount)
-	return res
-	`)
-
 	now := time.Now()
-	res, err := script.Run(r.client, []string{
+	res, err := currentStatsCmd.Run(r.client, []string{
 		base.AllQueues,
 		base.InProgressQueue,
 		base.ScheduledQueue,
@@ -173,6 +172,17 @@ func (r *RDB) CurrentStats() (*Stats, error) {
 	return stats, nil
 }

+var historicalStatsCmd = redis.NewScript(`
+local res = {}
+for _, key in ipairs(KEYS) do
+	local n = redis.call("GET", key)
+	if not n then
+	n = 0
+	end
+	table.insert(res, tonumber(n))
+end
+return res`)
+
 // HistoricalStats returns a list of stats from the last n days.
 func (r *RDB) HistoricalStats(n int) ([]*DailyStats, error) {
 	if n < 1 {
@@ -188,18 +198,7 @@ func (r *RDB) HistoricalStats(n int) ([]*DailyStats, error) {
 		keys = append(keys, base.ProcessedKey(ts))
 		keys = append(keys, base.FailureKey(ts))
 	}
-	script := redis.NewScript(`
-	local res = {}
-	for _, key in ipairs(KEYS) do
-	  local n = redis.call("GET", key)
-	  if not n then
-		n = 0
-	  end
-	  table.insert(res, tonumber(n))
-	end
-	return res
-	`)
-	res, err := script.Run(r.client, keys, len(keys)).Result()
+	res, err := historicalStatsCmd.Run(r.client, keys, len(keys)).Result()
 	if err != nil {
 		return nil, err
 	}
@@ -235,67 +234,46 @@ func (r *RDB) RedisInfo() (map[string]string, error) {
 	return info, nil
 }

+func reverse(x []string) {
+	for i := len(x)/2 - 1; i >= 0; i-- {
+		opp := len(x) - 1 - i
+		x[i], x[opp] = x[opp], x[i]
+	}
+}
+
+// Pagination specifies the page size and page number
+// for the list operation.
+type Pagination struct {
+	// Number of items in the page.
+	Size int
+
+	// Page number starting from zero.
+	Page int
+}
+
+func (p Pagination) start() int64 {
+	return int64(p.Size * p.Page)
+}
+
+func (p Pagination) stop() int64 {
+	return int64(p.Size*p.Page + p.Size - 1)
+}
+
 // ListEnqueued returns enqueued tasks that are ready to be processed.
-//
-// Queue names can be optionally passed to query only the specified queues.
-// If none are passed, it will query all queues.
-func (r *RDB) ListEnqueued(qnames ...string) ([]*EnqueuedTask, error) {
-	if len(qnames) == 0 {
-		return r.listAllEnqueued()
+func (r *RDB) ListEnqueued(qname string, pgn Pagination) ([]*EnqueuedTask, error) {
+	qkey := base.QueueKey(qname)
+	if !r.client.SIsMember(base.AllQueues, qkey).Val() {
+		return nil, fmt.Errorf("queue %q does not exist", qname)
 	}
-	return r.listEnqueued(qnames...)
-}
-
-func (r *RDB) listAllEnqueued() ([]*EnqueuedTask, error) {
-	script := redis.NewScript(`
-	local res = {}
-	local queues = redis.call("SMEMBERS", KEYS[1])
-	for _, qkey in ipairs(queues) do
-		local msgs = redis.call("LRANGE", qkey, 0, -1)
-		for _, msg in ipairs(msgs) do
-			table.insert(res, msg)
-		end
-	end
-	return res
-	`)
-	res, err := script.Run(r.client, []string{base.AllQueues}).Result()
+	// Note: Because we use LPUSH to redis list, we need to calculate the
+	// correct range and reverse the list to get the tasks with pagination.
+	stop := -pgn.start() - 1
+	start := -pgn.stop() - 1
+	data, err := r.client.LRange(qkey, start, stop).Result()
 	if err != nil {
 		return nil, err
 	}
-	data, err := cast.ToStringSliceE(res)
-	if err != nil {
-		return nil, err
-	}
-	return toEnqueuedTasks(data)
-}
-
-func (r *RDB) listEnqueued(qnames ...string) ([]*EnqueuedTask, error) {
-	script := redis.NewScript(`
-	local res = {}
-	for _, qkey in ipairs(KEYS) do
-		local msgs = redis.call("LRANGE", qkey, 0, -1)
-		for _, msg in ipairs(msgs) do
-			table.insert(res, msg)
-		end
-	end
-	return res
-	`)
-	var keys []string
-	for _, q := range qnames {
-		keys = append(keys, base.QueueKey(q))
-	}
-	res, err := script.Run(r.client, keys).Result()
-	if err != nil {
-		return nil, err
-	}
-	data, err := cast.ToStringSliceE(res)
-	if err != nil {
-		return nil, err
-	}
-	return toEnqueuedTasks(data)
-}
-
-func toEnqueuedTasks(data []string) ([]*EnqueuedTask, error) {
+	reverse(data)
 	var tasks []*EnqueuedTask
 	for _, s := range data {
 		var msg base.TaskMessage
@@ -314,11 +292,16 @@ func toEnqueuedTasks(data []string) ([]*EnqueuedTask, error) {
 }

 // ListInProgress returns all tasks that are currently being processed.
-func (r *RDB) ListInProgress() ([]*InProgressTask, error) {
-	data, err := r.client.LRange(base.InProgressQueue, 0, -1).Result()
+func (r *RDB) ListInProgress(pgn Pagination) ([]*InProgressTask, error) {
+	// Note: Because we use LPUSH to redis list, we need to calculate the
+	// correct range and reverse the list to get the tasks with pagination.
+	stop := -pgn.start() - 1
+	start := -pgn.stop() - 1
+	data, err := r.client.LRange(base.InProgressQueue, start, stop).Result()
 	if err != nil {
 		return nil, err
 	}
+	reverse(data)
 	var tasks []*InProgressTask
 	for _, s := range data {
 		var msg base.TaskMessage
@@ -337,8 +320,8 @@ func (r *RDB) ListInProgress() ([]*InProgressTask, error) {

 // ListScheduled returns all tasks that are scheduled to be processed
 // in the future.
-func (r *RDB) ListScheduled() ([]*ScheduledTask, error) {
-	data, err := r.client.ZRangeWithScores(base.ScheduledQueue, 0, -1).Result()
+func (r *RDB) ListScheduled(pgn Pagination) ([]*ScheduledTask, error) {
+	data, err := r.client.ZRangeWithScores(base.ScheduledQueue, pgn.start(), pgn.stop()).Result()
 	if err != nil {
 		return nil, err
 	}
@@ -368,8 +351,8 @@ func (r *RDB) ListScheduled() ([]*ScheduledTask, error) {

 // ListRetry returns all tasks that have failed before and willl be retried
 // in the future.
-func (r *RDB) ListRetry() ([]*RetryTask, error) {
-	data, err := r.client.ZRangeWithScores(base.RetryQueue, 0, -1).Result()
+func (r *RDB) ListRetry(pgn Pagination) ([]*RetryTask, error) {
+	data, err := r.client.ZRangeWithScores(base.RetryQueue, pgn.start(), pgn.stop()).Result()
 	if err != nil {
 		return nil, err
 	}
@@ -401,8 +384,8 @@ func (r *RDB) ListRetry() ([]*RetryTask, error) {
 }

 // ListDead returns all tasks that have exhausted its retry limit.
-func (r *RDB) ListDead() ([]*DeadTask, error) {
-	data, err := r.client.ZRangeWithScores(base.DeadQueue, 0, -1).Result()
+func (r *RDB) ListDead(pgn Pagination) ([]*DeadTask, error) {
+	data, err := r.client.ZRangeWithScores(base.DeadQueue, pgn.start(), pgn.stop()).Result()
 	if err != nil {
 		return nil, err
 	}
@@ -491,21 +474,21 @@ func (r *RDB) EnqueueAllDeadTasks() (int64, error) {
 	return r.removeAndEnqueueAll(base.DeadQueue)
 }

-func (r *RDB) removeAndEnqueue(zset, id string, score float64) (int64, error) {
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], ARGV[1], ARGV[1])
-	for _, msg in ipairs(msgs) do
-		local decoded = cjson.decode(msg)
-		if decoded["ID"] == ARGV[2] then
-			redis.call("ZREM", KEYS[1], msg)
-			local qkey = ARGV[3] .. decoded["Queue"]
-			redis.call("LPUSH", qkey, msg)
-			return 1
-		end
+var removeAndEnqueueCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], ARGV[1], ARGV[1])
+for _, msg in ipairs(msgs) do
+	local decoded = cjson.decode(msg)
+	if decoded["ID"] == ARGV[2] then
+		local qkey = ARGV[3] .. decoded["Queue"]
+		redis.call("LPUSH", qkey, msg)
+		redis.call("ZREM", KEYS[1], msg)
+		return 1
 	end
-	return 0
-	`)
-	res, err := script.Run(r.client, []string{zset}, score, id, base.QueuePrefix).Result()
+end
+return 0`)
+
+func (r *RDB) removeAndEnqueue(zset, id string, score float64) (int64, error) {
+	res, err := removeAndEnqueueCmd.Run(r.client, []string{zset}, score, id, base.QueuePrefix).Result()
 	if err != nil {
 		return 0, err
 	}
@@ -516,18 +499,18 @@ func (r *RDB) removeAndEnqueue(zset, id string, score float64) (int64, error) {
 	return n, nil
 }

+var removeAndEnqueueAllCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGE", KEYS[1], 0, -1)
+for _, msg in ipairs(msgs) do
+	local decoded = cjson.decode(msg)
+	local qkey = ARGV[1] .. decoded["Queue"]
+	redis.call("LPUSH", qkey, msg)
+	redis.call("ZREM", KEYS[1], msg)
+end
+return table.getn(msgs)`)
+
 func (r *RDB) removeAndEnqueueAll(zset string) (int64, error) {
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGE", KEYS[1], 0, -1)
-	for _, msg in ipairs(msgs) do
-		redis.call("ZREM", KEYS[1], msg)
-		local decoded = cjson.decode(msg)
-		local qkey = ARGV[1] .. decoded["Queue"]
-		redis.call("LPUSH", qkey, msg)
-	end
-	return table.getn(msgs)
-	`)
-	res, err := script.Run(r.client, []string{zset}, base.QueuePrefix).Result()
+	res, err := removeAndEnqueueAllCmd.Run(r.client, []string{zset}, base.QueuePrefix).Result()
 	if err != nil {
 		return 0, err
 	}
@@ -578,31 +561,31 @@ func (r *RDB) KillAllScheduledTasks() (int64, error) {
 	return r.removeAndKillAll(base.ScheduledQueue)
 }

-func (r *RDB) removeAndKill(zset, id string, score float64) (int64, error) {
-	// KEYS[1] -> ZSET to move task from (e.g., retry queue)
-	// KEYS[2] -> asynq:dead
-	// ARGV[1] -> score of the task to kill
-	// ARGV[2] -> id of the task to kill
-	// ARGV[3] -> current timestamp
-	// ARGV[4] -> cutoff timestamp (e.g., 90 days ago)
-	// ARGV[5] -> max number of tasks in dead queue (e.g., 100)
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], ARGV[1], ARGV[1])
-	for _, msg in ipairs(msgs) do
-		local decoded = cjson.decode(msg)
-		if decoded["ID"] == ARGV[2] then
-			redis.call("ZREM", KEYS[1], msg)
-			redis.call("ZADD", KEYS[2], ARGV[3], msg)
-			redis.call("ZREMRANGEBYSCORE", KEYS[2], "-inf", ARGV[4])
-			redis.call("ZREMRANGEBYRANK", KEYS[2], 0, -ARGV[5])
-			return 1
-		end
+// KEYS[1] -> ZSET to move task from (e.g., retry queue)
+// KEYS[2] -> asynq:dead
+// ARGV[1] -> score of the task to kill
+// ARGV[2] -> id of the task to kill
+// ARGV[3] -> current timestamp
+// ARGV[4] -> cutoff timestamp (e.g., 90 days ago)
+// ARGV[5] -> max number of tasks in dead queue (e.g., 100)
+var removeAndKillCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], ARGV[1], ARGV[1])
+for _, msg in ipairs(msgs) do
+	local decoded = cjson.decode(msg)
+	if decoded["ID"] == ARGV[2] then
+		redis.call("ZREM", KEYS[1], msg)
+		redis.call("ZADD", KEYS[2], ARGV[3], msg)
+		redis.call("ZREMRANGEBYSCORE", KEYS[2], "-inf", ARGV[4])
+		redis.call("ZREMRANGEBYRANK", KEYS[2], 0, -ARGV[5])
+		return 1
 	end
-	return 0
-	`)
+end
+return 0`)
+
+func (r *RDB) removeAndKill(zset, id string, score float64) (int64, error) {
 	now := time.Now()
 	limit := now.AddDate(0, 0, -deadExpirationInDays).Unix() // 90 days ago
-	res, err := script.Run(r.client,
+	res, err := removeAndKillCmd.Run(r.client,
 		[]string{zset, base.DeadQueue},
 		score, id, now.Unix(), limit, maxDeadTasks).Result()
 	if err != nil {
@@ -615,25 +598,25 @@ func (r *RDB) removeAndKill(zset, id string, score float64) (int64, error) {
 	return n, nil
 }

+// KEYS[1] -> ZSET to move task from (e.g., retry queue)
+// KEYS[2] -> asynq:dead
+// ARGV[1] -> current timestamp
+// ARGV[2] -> cutoff timestamp (e.g., 90 days ago)
+// ARGV[3] -> max number of tasks in dead queue (e.g., 100)
+var removeAndKillAllCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGE", KEYS[1], 0, -1)
+for _, msg in ipairs(msgs) do
+	redis.call("ZADD", KEYS[2], ARGV[1], msg)
+	redis.call("ZREM", KEYS[1], msg)
+	redis.call("ZREMRANGEBYSCORE", KEYS[2], "-inf", ARGV[2])
+	redis.call("ZREMRANGEBYRANK", KEYS[2], 0, -ARGV[3])
+end
+return table.getn(msgs)`)
+
 func (r *RDB) removeAndKillAll(zset string) (int64, error) {
-	// KEYS[1] -> ZSET to move task from (e.g., retry queue)
-	// KEYS[2] -> asynq:dead
-	// ARGV[1] -> current timestamp
-	// ARGV[2] -> cutoff timestamp (e.g., 90 days ago)
-	// ARGV[3] -> max number of tasks in dead queue (e.g., 100)
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGE", KEYS[1], 0, -1)
-	for _, msg in ipairs(msgs) do
-		redis.call("ZREM", KEYS[1], msg)
-		redis.call("ZADD", KEYS[2], ARGV[1], msg)
-		redis.call("ZREMRANGEBYSCORE", KEYS[2], "-inf", ARGV[2])
-		redis.call("ZREMRANGEBYRANK", KEYS[2], 0, -ARGV[3])
-	end
-	return table.getn(msgs)
-	`)
 	now := time.Now()
 	limit := now.AddDate(0, 0, -deadExpirationInDays).Unix() // 90 days ago
-	res, err := script.Run(r.client, []string{zset, base.DeadQueue},
+	res, err := removeAndKillAllCmd.Run(r.client, []string{zset, base.DeadQueue},
 		now.Unix(), limit, maxDeadTasks).Result()
 	if err != nil {
 		return 0, err
@@ -666,19 +649,19 @@ func (r *RDB) DeleteScheduledTask(id xid.ID, score int64) error {
 	return r.deleteTask(base.ScheduledQueue, id.String(), float64(score))
 }

-func (r *RDB) deleteTask(zset, id string, score float64) error {
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], ARGV[1], ARGV[1])
-	for _, msg in ipairs(msgs) do
-		local decoded = cjson.decode(msg)
-		if decoded["ID"] == ARGV[2] then
-			redis.call("ZREM", KEYS[1], msg)
-			return 1
-		end
+var deleteTaskCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], ARGV[1], ARGV[1])
+for _, msg in ipairs(msgs) do
+	local decoded = cjson.decode(msg)
+	if decoded["ID"] == ARGV[2] then
+		redis.call("ZREM", KEYS[1], msg)
+		return 1
 	end
-	return 0
-	`)
-	res, err := script.Run(r.client, []string{zset}, score, id).Result()
+end
+return 0`)
+
+func (r *RDB) deleteTask(zset, id string, score float64) error {
+	res, err := deleteTaskCmd.Run(r.client, []string{zset}, score, id).Result()
 	if err != nil {
 		return err
 	}
@@ -725,6 +708,27 @@ func (e *ErrQueueNotEmpty) Error() string {
 	return fmt.Sprintf("queue %q is not empty", e.qname)
 }

+// Skip checking whether queue is empty before removing.
+var removeQueueForceCmd = redis.NewScript(`
+local n = redis.call("SREM", KEYS[1], KEYS[2])
+if n == 0 then
+	return redis.error_reply("LIST NOT FOUND")
+end
+redis.call("DEL", KEYS[2])
+return redis.status_reply("OK")`)
+
+// Checks whether queue is empty before removing.
+var removeQueueCmd = redis.NewScript(`
+local l = redis.call("LLEN", KEYS[2]) if l > 0 then
+	return redis.error_reply("LIST NOT EMPTY")
+end
+local n = redis.call("SREM", KEYS[1], KEYS[2])
+if n == 0 then
+	return redis.error_reply("LIST NOT FOUND")
+end
+redis.call("DEL", KEYS[2])
+return redis.status_reply("OK")`)
+
 // RemoveQueue removes the specified queue.
 //
 // If force is set to true, it will remove the queue regardless
@@ -734,27 +738,9 @@ func (e *ErrQueueNotEmpty) Error() string {
 func (r *RDB) RemoveQueue(qname string, force bool) error {
 	var script *redis.Script
 	if force {
-		script = redis.NewScript(`
-		local n = redis.call("SREM", KEYS[1], KEYS[2])
-		if n == 0 then
-			return redis.error_reply("LIST NOT FOUND")
-		end
-		redis.call("DEL", KEYS[2])
-		return redis.status_reply("OK")
-		`)
+		script = removeQueueForceCmd
 	} else {
-		script = redis.NewScript(`
-		local l = redis.call("LLEN", KEYS[2])
-		if l > 0 then
-			return redis.error_reply("LIST NOT EMPTY")
-		end
-		local n = redis.call("SREM", KEYS[1], KEYS[2])
-		if n == 0 then
-			return redis.error_reply("LIST NOT FOUND")
-		end
-		redis.call("DEL", KEYS[2])
-		return redis.status_reply("OK")
-		`)
+		script = removeQueueCmd
 	}
 	err := script.Run(r.client,
 		[]string{base.AllQueues, base.QueueKey(qname)},
@@ -771,3 +757,40 @@ func (r *RDB) RemoveQueue(qname string, force bool) error {
 	}
 	return nil
 }
+
+// Note: Script also removes stale keys.
+var listProcessesCmd = redis.NewScript(`
+local res = {}
+local now = tonumber(ARGV[1])
+local keys = redis.call("ZRANGEBYSCORE", KEYS[1], now, "+inf")
+for _, key in ipairs(keys) do
+	local ps = redis.call("GET", key)
+	if ps then
+		table.insert(res, ps)
+	end  
+end
+redis.call("ZREMRANGEBYSCORE", KEYS[1], "-inf", now-1)
+return res`)
+
+// ListProcesses returns the list of process statuses.
+func (r *RDB) ListProcesses() ([]*base.ProcessInfo, error) {
+	res, err := listProcessesCmd.Run(r.client,
+		[]string{base.AllProcesses}, time.Now().UTC().Unix()).Result()
+	if err != nil {
+		return nil, err
+	}
+	data, err := cast.ToStringSliceE(res)
+	if err != nil {
+		return nil, err
+	}
+	var processes []*base.ProcessInfo
+	for _, s := range data {
+		var ps base.ProcessInfo
+		err := json.Unmarshal([]byte(s), &ps)
+		if err != nil {
+			continue // skip bad data
+		}
+		processes = append(processes, &ps)
+	}
+	return processes, nil
+}
--- a/internal/rdb/inspect_test.go
+++ b/internal/rdb/inspect_test.go
@@ -5,6 +5,7 @@
 package rdb

 import (
+	"fmt"
 	"sort"
 	"testing"
 	"time"
@@ -231,25 +232,24 @@ func TestListEnqueued(t *testing.T) {
 	t1 := &EnqueuedTask{ID: m1.ID, Type: m1.Type, Payload: m1.Payload, Queue: m1.Queue}
 	t2 := &EnqueuedTask{ID: m2.ID, Type: m2.Type, Payload: m2.Payload, Queue: m2.Queue}
 	t3 := &EnqueuedTask{ID: m3.ID, Type: m3.Type, Payload: m3.Payload, Queue: m3.Queue}
-	t4 := &EnqueuedTask{ID: m4.ID, Type: m4.Type, Payload: m4.Payload, Queue: m4.Queue}
 	tests := []struct {
 		enqueued map[string][]*base.TaskMessage
-		qnames   []string
+		qname    string
 		want     []*EnqueuedTask
 	}{
 		{
 			enqueued: map[string][]*base.TaskMessage{
 				base.DefaultQueueName: {m1, m2},
 			},
-			qnames: []string{},
-			want:   []*EnqueuedTask{t1, t2},
+			qname: base.DefaultQueueName,
+			want:  []*EnqueuedTask{t1, t2},
 		},
 		{
 			enqueued: map[string][]*base.TaskMessage{
 				base.DefaultQueueName: {},
 			},
-			qnames: []string{},
-			want:   []*EnqueuedTask{},
+			qname: base.DefaultQueueName,
+			want:  []*EnqueuedTask{},
 		},
 		{
 			enqueued: map[string][]*base.TaskMessage{
@@ -257,8 +257,8 @@ func TestListEnqueued(t *testing.T) {
 				"critical":            {m3},
 				"low":                 {m4},
 			},
-			qnames: []string{},
-			want:   []*EnqueuedTask{t1, t2, t3, t4},
+			qname: base.DefaultQueueName,
+			want:  []*EnqueuedTask{t1, t2},
 		},
 		{
 			enqueued: map[string][]*base.TaskMessage{
@@ -266,17 +266,8 @@ func TestListEnqueued(t *testing.T) {
 				"critical":            {m3},
 				"low":                 {m4},
 			},
-			qnames: []string{"critical"},
-			want:   []*EnqueuedTask{t3},
-		},
-		{
-			enqueued: map[string][]*base.TaskMessage{
-				base.DefaultQueueName: {m1, m2},
-				"critical":            {m3},
-				"low":                 {m4},
-			},
-			qnames: []string{"critical", "low"},
-			want:   []*EnqueuedTask{t3, t4},
+			qname: "critical",
+			want:  []*EnqueuedTask{t3},
 		},
 	}

@@ -286,9 +277,10 @@ func TestListEnqueued(t *testing.T) {
 			h.SeedEnqueuedQueue(t, r.client, msgs, qname)
 		}

-		got, err := r.ListEnqueued(tc.qnames...)
+		got, err := r.ListEnqueued(tc.qname, Pagination{Size: 20, Page: 0})
+		op := fmt.Sprintf("r.ListEnqueued(%q, Pagination{Size: 20, Page: 0})", tc.qname)
 		if err != nil {
-			t.Errorf("r.ListEnqueued() = %v, %v, want %v, nil", got, err, tc.want)
+			t.Errorf("%s = %v, %v, want %v, nil", op, got, err, tc.want)
 			continue
 		}
 		sortOpt := cmp.Transformer("SortMsg", func(in []*EnqueuedTask) []*EnqueuedTask {
@@ -299,11 +291,76 @@ func TestListEnqueued(t *testing.T) {
 			return out
 		})
 		if diff := cmp.Diff(tc.want, got, sortOpt); diff != "" {
-			t.Errorf("r.ListEnqueued() = %v, %v, want %v, nil; (-want, +got)\n%s", got, err, tc.want, diff)
+			t.Errorf("%s = %v, %v, want %v, nil; (-want, +got)\n%s", op, got, err, tc.want, diff)
 			continue
 		}
 	}
 }
+func TestListEnqueuedPagination(t *testing.T) {
+	r := setup(t)
+	var msgs []*base.TaskMessage
+	for i := 0; i < 100; i++ {
+		msg := h.NewTaskMessage(fmt.Sprintf("task %d", i), nil)
+		msgs = append(msgs, msg)
+	}
+	// create 100 tasks in default queue
+	h.SeedEnqueuedQueue(t, r.client, msgs)
+
+	msgs = []*base.TaskMessage(nil) // empty list
+	for i := 0; i < 100; i++ {
+		msg := h.NewTaskMessage(fmt.Sprintf("custom %d", i), nil)
+		msgs = append(msgs, msg)
+	}
+	// create 100 tasks in custom queue
+	h.SeedEnqueuedQueue(t, r.client, msgs, "custom")
+
+	tests := []struct {
+		desc      string
+		qname     string
+		page      int
+		size      int
+		wantSize  int
+		wantFirst string
+		wantLast  string
+	}{
+		{"first page", "default", 0, 20, 20, "task 0", "task 19"},
+		{"second page", "default", 1, 20, 20, "task 20", "task 39"},
+		{"different page size", "default", 2, 30, 30, "task 60", "task 89"},
+		{"last page", "default", 3, 30, 10, "task 90", "task 99"},
+		{"out of range", "default", 4, 30, 0, "", ""},
+		{"second page with custom queue", "custom", 1, 20, 20, "custom 20", "custom 39"},
+	}
+
+	for _, tc := range tests {
+		got, err := r.ListEnqueued(tc.qname, Pagination{Size: tc.size, Page: tc.page})
+		op := fmt.Sprintf("r.ListEnqueued(%q, Pagination{Size: %d, Page: %d})", tc.qname, tc.size, tc.page)
+		if err != nil {
+			t.Errorf("%s; %s returned error %v", tc.desc, op, err)
+			continue
+		}
+
+		if len(got) != tc.wantSize {
+			t.Errorf("%s; %s returned a list of size %d, want %d", tc.desc, op, len(got), tc.wantSize)
+			continue
+		}
+
+		if tc.wantSize == 0 {
+			continue
+		}
+
+		first := got[0]
+		if first.Type != tc.wantFirst {
+			t.Errorf("%s; %s returned a list with first message %q, want %q",
+				tc.desc, op, first.Type, tc.wantFirst)
+		}
+
+		last := got[len(got)-1]
+		if last.Type != tc.wantLast {
+			t.Errorf("%s; %s returned a list with the last message %q, want %q",
+				tc.desc, op, last.Type, tc.wantLast)
+		}
+	}
+}

 func TestListInProgress(t *testing.T) {
 	r := setup(t)
@@ -330,9 +387,10 @@ func TestListInProgress(t *testing.T) {
 		h.FlushDB(t, r.client) // clean up db before each test case
 		h.SeedInProgressQueue(t, r.client, tc.inProgress)

-		got, err := r.ListInProgress()
+		got, err := r.ListInProgress(Pagination{Size: 20, Page: 0})
+		op := "r.ListInProgress(Pagination{Size: 20, Page: 0})"
 		if err != nil {
-			t.Errorf("r.ListInProgress() = %v, %v, want %v, nil", got, err, tc.want)
+			t.Errorf("%s = %v, %v, want %v, nil", op, got, err, tc.want)
 			continue
 		}
 		sortOpt := cmp.Transformer("SortMsg", func(in []*InProgressTask) []*InProgressTask {
@@ -343,12 +401,67 @@ func TestListInProgress(t *testing.T) {
 			return out
 		})
 		if diff := cmp.Diff(tc.want, got, sortOpt); diff != "" {
-			t.Errorf("r.ListInProgress() = %v, %v, want %v, nil; (-want, +got)\n%s", got, err, tc.want, diff)
+			t.Errorf("%s = %v, %v, want %v, nil; (-want, +got)\n%s", op, got, err, tc.want, diff)
 			continue
 		}
 	}
 }

+func TestListInProgressPagination(t *testing.T) {
+	r := setup(t)
+	var msgs []*base.TaskMessage
+	for i := 0; i < 100; i++ {
+		msg := h.NewTaskMessage(fmt.Sprintf("task %d", i), nil)
+		msgs = append(msgs, msg)
+	}
+	h.SeedInProgressQueue(t, r.client, msgs)
+
+	tests := []struct {
+		desc      string
+		page      int
+		size      int
+		wantSize  int
+		wantFirst string
+		wantLast  string
+	}{
+		{"first page", 0, 20, 20, "task 0", "task 19"},
+		{"second page", 1, 20, 20, "task 20", "task 39"},
+		{"different page size", 2, 30, 30, "task 60", "task 89"},
+		{"last page", 3, 30, 10, "task 90", "task 99"},
+		{"out of range", 4, 30, 0, "", ""},
+	}
+
+	for _, tc := range tests {
+		got, err := r.ListInProgress(Pagination{Size: tc.size, Page: tc.page})
+		op := fmt.Sprintf("r.ListInProgress(Pagination{Size: %d, Page: %d})", tc.size, tc.page)
+		if err != nil {
+			t.Errorf("%s; %s returned error %v", tc.desc, op, err)
+			continue
+		}
+
+		if len(got) != tc.wantSize {
+			t.Errorf("%s; %s returned list of size %d, want %d", tc.desc, op, len(got), tc.wantSize)
+			continue
+		}
+
+		if tc.wantSize == 0 {
+			continue
+		}
+
+		first := got[0]
+		if first.Type != tc.wantFirst {
+			t.Errorf("%s; %s returned a list with first message %q, want %q",
+				tc.desc, op, first.Type, tc.wantFirst)
+		}
+
+		last := got[len(got)-1]
+		if last.Type != tc.wantLast {
+			t.Errorf("%s; %s returned a list with the last message %q, want %q",
+				tc.desc, op, last.Type, tc.wantLast)
+		}
+	}
+}
+
 func TestListScheduled(t *testing.T) {
 	r := setup(t)
 	m1 := h.NewTaskMessage("send_email", map[string]interface{}{"subject": "hello"})
@@ -379,9 +492,10 @@ func TestListScheduled(t *testing.T) {
 		h.FlushDB(t, r.client) // clean up db before each test case
 		h.SeedScheduledQueue(t, r.client, tc.scheduled)

-		got, err := r.ListScheduled()
+		got, err := r.ListScheduled(Pagination{Size: 20, Page: 0})
+		op := "r.ListScheduled(Pagination{Size: 20, Page: 0})"
 		if err != nil {
-			t.Errorf("r.ListScheduled() = %v, %v, want %v, nil", got, err, tc.want)
+			t.Errorf("%s = %v, %v, want %v, nil", op, got, err, tc.want)
 			continue
 		}
 		sortOpt := cmp.Transformer("SortMsg", func(in []*ScheduledTask) []*ScheduledTask {
@@ -392,12 +506,68 @@ func TestListScheduled(t *testing.T) {
 			return out
 		})
 		if diff := cmp.Diff(tc.want, got, sortOpt, timeCmpOpt); diff != "" {
-			t.Errorf("r.ListScheduled() = %v, %v, want %v, nil; (-want, +got)\n%s", got, err, tc.want, diff)
+			t.Errorf("%s = %v, %v, want %v, nil; (-want, +got)\n%s", op, got, err, tc.want, diff)
 			continue
 		}
 	}
 }

+func TestListScheduledPagination(t *testing.T) {
+	r := setup(t)
+	// create 100 tasks with an increasing number of wait time.
+	for i := 0; i < 100; i++ {
+		msg := h.NewTaskMessage(fmt.Sprintf("task %d", i), nil)
+		if err := r.Schedule(msg, time.Now().Add(time.Duration(i)*time.Second)); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	tests := []struct {
+		desc      string
+		page      int
+		size      int
+		wantSize  int
+		wantFirst string
+		wantLast  string
+	}{
+		{"first page", 0, 20, 20, "task 0", "task 19"},
+		{"second page", 1, 20, 20, "task 20", "task 39"},
+		{"different page size", 2, 30, 30, "task 60", "task 89"},
+		{"last page", 3, 30, 10, "task 90", "task 99"},
+		{"out of range", 4, 30, 0, "", ""},
+	}
+
+	for _, tc := range tests {
+		got, err := r.ListScheduled(Pagination{Size: tc.size, Page: tc.page})
+		op := fmt.Sprintf("r.ListScheduled(Pagination{Size: %d, Page: %d})", tc.size, tc.page)
+		if err != nil {
+			t.Errorf("%s; %s returned error %v", tc.desc, op, err)
+			continue
+		}
+
+		if len(got) != tc.wantSize {
+			t.Errorf("%s; %s returned list of size %d, want %d", tc.desc, op, len(got), tc.wantSize)
+			continue
+		}
+
+		if tc.wantSize == 0 {
+			continue
+		}
+
+		first := got[0]
+		if first.Type != tc.wantFirst {
+			t.Errorf("%s; %s returned a list with first message %q, want %q",
+				tc.desc, op, first.Type, tc.wantFirst)
+		}
+
+		last := got[len(got)-1]
+		if last.Type != tc.wantLast {
+			t.Errorf("%s; %s returned a list with the last message %q, want %q",
+				tc.desc, op, last.Type, tc.wantLast)
+		}
+	}
+}
+
 func TestListRetry(t *testing.T) {
 	r := setup(t)
 	m1 := &base.TaskMessage{
@@ -464,9 +634,10 @@ func TestListRetry(t *testing.T) {
 		h.FlushDB(t, r.client) // clean up db before each test case
 		h.SeedRetryQueue(t, r.client, tc.retry)

-		got, err := r.ListRetry()
+		got, err := r.ListRetry(Pagination{Size: 20, Page: 0})
+		op := "r.ListRetry(Pagination{Size: 20, Page: 0})"
 		if err != nil {
-			t.Errorf("r.ListRetry() = %v, %v, want %v, nil", got, err, tc.want)
+			t.Errorf("%s = %v, %v, want %v, nil", op, got, err, tc.want)
 			continue
 		}
 		sortOpt := cmp.Transformer("SortMsg", func(in []*RetryTask) []*RetryTask {
@@ -478,12 +649,68 @@ func TestListRetry(t *testing.T) {
 		})

 		if diff := cmp.Diff(tc.want, got, sortOpt, timeCmpOpt); diff != "" {
-			t.Errorf("r.ListRetry() = %v, %v, want %v, nil; (-want, +got)\n%s", got, err, tc.want, diff)
+			t.Errorf("%s = %v, %v, want %v, nil; (-want, +got)\n%s", op, got, err, tc.want, diff)
 			continue
 		}
 	}
 }

+func TestListRetryPagination(t *testing.T) {
+	r := setup(t)
+	// create 100 tasks with an increasing number of wait time.
+	for i := 0; i < 100; i++ {
+		msg := h.NewTaskMessage(fmt.Sprintf("task %d", i), nil)
+		if err := r.Retry(msg, time.Now().Add(time.Duration(i)*time.Second), "error"); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	tests := []struct {
+		desc      string
+		page      int
+		size      int
+		wantSize  int
+		wantFirst string
+		wantLast  string
+	}{
+		{"first page", 0, 20, 20, "task 0", "task 19"},
+		{"second page", 1, 20, 20, "task 20", "task 39"},
+		{"different page size", 2, 30, 30, "task 60", "task 89"},
+		{"last page", 3, 30, 10, "task 90", "task 99"},
+		{"out of range", 4, 30, 0, "", ""},
+	}
+
+	for _, tc := range tests {
+		got, err := r.ListRetry(Pagination{Size: tc.size, Page: tc.page})
+		op := fmt.Sprintf("r.ListRetry(Pagination{Size: %d, Page: %d})", tc.size, tc.page)
+		if err != nil {
+			t.Errorf("%s; %s returned error %v", tc.desc, op, err)
+			continue
+		}
+
+		if len(got) != tc.wantSize {
+			t.Errorf("%s; %s returned list of size %d, want %d", tc.desc, op, len(got), tc.wantSize)
+			continue
+		}
+
+		if tc.wantSize == 0 {
+			continue
+		}
+
+		first := got[0]
+		if first.Type != tc.wantFirst {
+			t.Errorf("%s; %s returned a list with first message %q, want %q",
+				tc.desc, op, first.Type, tc.wantFirst)
+		}
+
+		last := got[len(got)-1]
+		if last.Type != tc.wantLast {
+			t.Errorf("%s; %s returned a list with the last message %q, want %q",
+				tc.desc, op, last.Type, tc.wantLast)
+		}
+	}
+}
+
 func TestListDead(t *testing.T) {
 	r := setup(t)
 	m1 := &base.TaskMessage{
@@ -542,9 +769,10 @@ func TestListDead(t *testing.T) {
 		h.FlushDB(t, r.client) // clean up db before each test case
 		h.SeedDeadQueue(t, r.client, tc.dead)

-		got, err := r.ListDead()
+		got, err := r.ListDead(Pagination{Size: 20, Page: 0})
+		op := "r.ListDead(Pagination{Size: 20, Page: 0})"
 		if err != nil {
-			t.Errorf("r.ListDead() = %v, %v, want %v, nil", got, err, tc.want)
+			t.Errorf("%s = %v, %v, want %v, nil", op, got, err, tc.want)
 			continue
 		}
 		sortOpt := cmp.Transformer("SortMsg", func(in []*DeadTask) []*DeadTask {
@@ -555,12 +783,67 @@ func TestListDead(t *testing.T) {
 			return out
 		})
 		if diff := cmp.Diff(tc.want, got, sortOpt, timeCmpOpt); diff != "" {
-			t.Errorf("r.ListDead() = %v, %v, want %v, nil; (-want, +got)\n%s", got, err, tc.want, diff)
+			t.Errorf("%s = %v, %v, want %v, nil; (-want, +got)\n%s", op, got, err, tc.want, diff)
 			continue
 		}
 	}
 }

+func TestListDeadPagination(t *testing.T) {
+	r := setup(t)
+	var entries []h.ZSetEntry
+	for i := 0; i < 100; i++ {
+		msg := h.NewTaskMessage(fmt.Sprintf("task %d", i), nil)
+		entries = append(entries, h.ZSetEntry{Msg: msg, Score: float64(i)})
+	}
+	h.SeedDeadQueue(t, r.client, entries)
+
+	tests := []struct {
+		desc      string
+		page      int
+		size      int
+		wantSize  int
+		wantFirst string
+		wantLast  string
+	}{
+		{"first page", 0, 20, 20, "task 0", "task 19"},
+		{"second page", 1, 20, 20, "task 20", "task 39"},
+		{"different page size", 2, 30, 30, "task 60", "task 89"},
+		{"last page", 3, 30, 10, "task 90", "task 99"},
+		{"out of range", 4, 30, 0, "", ""},
+	}
+
+	for _, tc := range tests {
+		got, err := r.ListDead(Pagination{Size: tc.size, Page: tc.page})
+		op := fmt.Sprintf("r.ListDead(Pagination{Size: %d, Page: %d})", tc.size, tc.page)
+		if err != nil {
+			t.Errorf("%s; %s returned error %v", tc.desc, op, err)
+			continue
+		}
+
+		if len(got) != tc.wantSize {
+			t.Errorf("%s; %s returned list of size %d, want %d", tc.desc, op, len(got), tc.wantSize)
+			continue
+		}
+
+		if tc.wantSize == 0 {
+			continue
+		}
+
+		first := got[0]
+		if first.Type != tc.wantFirst {
+			t.Errorf("%s; %s returned a list with first message %q, want %q",
+				tc.desc, op, first.Type, tc.wantFirst)
+		}
+
+		last := got[len(got)-1]
+		if last.Type != tc.wantLast {
+			t.Errorf("%s; %s returned a list with the last message %q, want %q",
+				tc.desc, op, last.Type, tc.wantLast)
+		}
+	}
+}
+
 var timeCmpOpt = cmpopts.EquateApproxTime(time.Second)

 func TestEnqueueDeadTask(t *testing.T) {
@@ -1767,3 +2050,56 @@ func TestRemoveQueueError(t *testing.T) {
 		}
 	}
 }
+
+func TestListProcesses(t *testing.T) {
+	r := setup(t)
+
+	ps1 := &base.ProcessInfo{
+		Concurrency:       10,
+		Queues:            map[string]int{"default": 1},
+		Host:              "do.droplet1",
+		PID:               1234,
+		State:             "running",
+		Started:           time.Now().Add(-time.Hour),
+		ActiveWorkerCount: 5,
+	}
+
+	ps2 := &base.ProcessInfo{
+		Concurrency:       20,
+		Queues:            map[string]int{"email": 1},
+		Host:              "do.droplet2",
+		PID:               9876,
+		State:             "stopped",
+		Started:           time.Now().Add(-2 * time.Hour),
+		ActiveWorkerCount: 20,
+	}
+
+	tests := []struct {
+		processes []*base.ProcessInfo
+	}{
+		{processes: []*base.ProcessInfo{}},
+		{processes: []*base.ProcessInfo{ps1}},
+		{processes: []*base.ProcessInfo{ps1, ps2}},
+	}
+
+	ignoreOpt := cmpopts.IgnoreUnexported(base.ProcessInfo{})
+
+	for _, tc := range tests {
+		h.FlushDB(t, r.client)
+
+		for _, ps := range tc.processes {
+			if err := r.WriteProcessInfo(ps, 5*time.Second); err != nil {
+				t.Fatal(err)
+			}
+		}
+
+		got, err := r.ListProcesses()
+		if err != nil {
+			t.Errorf("r.ListProcesses returned an error: %v", err)
+		}
+		if diff := cmp.Diff(tc.processes, got, h.SortProcessInfoOpt, ignoreOpt); diff != "" {
+			t.Errorf("r.ListProcesses returned %v, want %v; (-want,+got)\n%s",
+				got, tc.processes, diff)
+		}
+	}
+}
--- a/internal/rdb/rdb.go
+++ b/internal/rdb/rdb.go
@@ -41,6 +41,14 @@ func (r *RDB) Close() error {
 	return r.client.Close()
 }

+// KEYS[1] -> asynq:queues:<qname>
+// KEYS[2] -> asynq:queues
+// ARGV[1] -> task message data
+var enqueueCmd = redis.NewScript(`
+redis.call("LPUSH", KEYS[1], ARGV[1])
+redis.call("SADD", KEYS[2], KEYS[1])
+return 1`)
+
 // Enqueue inserts the given task to the tail of the queue.
 func (r *RDB) Enqueue(msg *base.TaskMessage) error {
 	bytes, err := json.Marshal(msg)
@@ -48,23 +56,18 @@ func (r *RDB) Enqueue(msg *base.TaskMessage) error {
 		return err
 	}
 	key := base.QueueKey(msg.Queue)
-	script := redis.NewScript(`
-	redis.call("LPUSH", KEYS[1], ARGV[1])
-	redis.call("SADD", KEYS[2], KEYS[1])
-	return 1
-	`)
-	return script.Run(r.client, []string{key, base.AllQueues}, string(bytes)).Err()
+	return enqueueCmd.Run(r.client, []string{key, base.AllQueues}, bytes).Err()
 }

-// Dequeue queries given queues in order and pops a task message if there
-// is one and returns it. If all queues are empty, ErrNoProcessableTask
-// error is returned.
+// Dequeue queries given queues in order and pops a task message if there is one and returns it.
+// If all queues are empty, ErrNoProcessableTask error is returned.
 func (r *RDB) Dequeue(qnames ...string) (*base.TaskMessage, error) {
 	var data string
 	var err error
 	if len(qnames) == 1 {
 		data, err = r.dequeueSingle(base.QueueKey(qnames[0]))
 	} else {
+		// TODO(hibiken): Take keys are argument and don't compute every time
 		var keys []string
 		for _, q := range qnames {
 			keys = append(keys, base.QueueKey(q))
@@ -90,73 +93,75 @@ func (r *RDB) dequeueSingle(queue string) (data string, err error) {
 	return r.client.BRPopLPush(queue, base.InProgressQueue, time.Second).Result()
 }

+// KEYS[1] -> asynq:in_progress
+// ARGV    -> List of queues to query in order
+var dequeueCmd = redis.NewScript(`
+local res
+for _, qkey in ipairs(ARGV) do
+	res = redis.call("RPOPLPUSH", qkey, KEYS[1])
+	if res then
+		return res
+	end
+end
+return res`)
+
 func (r *RDB) dequeue(queues ...string) (data string, err error) {
 	var args []interface{}
 	for _, qkey := range queues {
 		args = append(args, qkey)
 	}
-	script := redis.NewScript(`
-	local res
-	for _, qkey in ipairs(ARGV) do
-		res = redis.call("RPOPLPUSH", qkey, KEYS[1])
-		if res then
-			return res
-		end
-	end
-	return res
-	`)
-	res, err := script.Run(r.client, []string{base.InProgressQueue}, args...).Result()
+	res, err := dequeueCmd.Run(r.client, []string{base.InProgressQueue}, args...).Result()
 	if err != nil {
 		return "", err
 	}
 	return cast.ToStringE(res)
 }

+// KEYS[1] -> asynq:in_progress
+// KEYS[2] -> asynq:processed:<yyyy-mm-dd>
+// ARGV[1] -> base.TaskMessage value
+// ARGV[2] -> stats expiration timestamp
+// Note: LREM count ZERO means "remove all elements equal to val"
+var doneCmd = redis.NewScript(`
+redis.call("LREM", KEYS[1], 0, ARGV[1]) 
+local n = redis.call("INCR", KEYS[2])
+if tonumber(n) == 1 then
+	redis.call("EXPIREAT", KEYS[2], ARGV[2])
+end
+return redis.status_reply("OK")
+`)
+
 // Done removes the task from in-progress queue to mark the task as done.
 func (r *RDB) Done(msg *base.TaskMessage) error {
 	bytes, err := json.Marshal(msg)
 	if err != nil {
 		return err
 	}
-	// Note: LREM count ZERO means "remove all elements equal to val"
-	// KEYS[1] -> asynq:in_progress
-	// KEYS[2] -> asynq:processed:<yyyy-mm-dd>
-	// ARGV[1] -> base.TaskMessage value
-	// ARGV[2] -> stats expiration timestamp
-	script := redis.NewScript(`
-	redis.call("LREM", KEYS[1], 0, ARGV[1]) 
-	local n = redis.call("INCR", KEYS[2])
-	if tonumber(n) == 1 then
-		redis.call("EXPIREAT", KEYS[2], ARGV[2])
-	end
-	return redis.status_reply("OK")
-	`)
 	now := time.Now()
 	processedKey := base.ProcessedKey(now)
 	expireAt := now.Add(statsTTL)
-	return script.Run(r.client,
+	return doneCmd.Run(r.client,
 		[]string{base.InProgressQueue, processedKey},
-		string(bytes), expireAt.Unix()).Err()
+		bytes, expireAt.Unix()).Err()
 }

-// Requeue moves the task from in-progress queue to the default
-// queue.
+// KEYS[1] -> asynq:in_progress
+// KEYS[2] -> asynq:queues:<qname>
+// ARGV[1] -> base.TaskMessage value
+// Note: Use RPUSH to push to the head of the queue.
+var requeueCmd = redis.NewScript(`
+redis.call("LREM", KEYS[1], 0, ARGV[1])
+redis.call("RPUSH", KEYS[2], ARGV[1])
+return redis.status_reply("OK")`)
+
+// Requeue moves the task from in-progress queue to the specified queue.
 func (r *RDB) Requeue(msg *base.TaskMessage) error {
 	bytes, err := json.Marshal(msg)
 	if err != nil {
 		return err
 	}
-	// Note: Use RPUSH to push to the head of the queue.
-	// KEYS[1] -> asynq:in_progress
-	// KEYS[2] -> asynq:queues:default
-	// ARGV[1] -> base.TaskMessage value
-	script := redis.NewScript(`
-	redis.call("LREM", KEYS[1], 0, ARGV[1])
-	redis.call("RPUSH", KEYS[2], ARGV[1])
-	return redis.status_reply("OK")
-	`)
-	return script.Run(r.client,
-		[]string{base.InProgressQueue, base.DefaultQueue},
+	return requeueCmd.Run(r.client,
+		[]string{base.InProgressQueue, base.QueueKey(msg.Queue)},
 		string(bytes)).Err()
 }

@@ -171,6 +176,27 @@ func (r *RDB) Schedule(msg *base.TaskMessage, processAt time.Time) error {
 		&redis.Z{Member: string(bytes), Score: score}).Err()
 }

+// KEYS[1] -> asynq:in_progress
+// KEYS[2] -> asynq:retry
+// KEYS[3] -> asynq:processed:<yyyy-mm-dd>
+// KEYS[4] -> asynq:failure:<yyyy-mm-dd>
+// ARGV[1] -> base.TaskMessage value to remove from base.InProgressQueue queue
+// ARGV[2] -> base.TaskMessage value to add to Retry queue
+// ARGV[3] -> retry_at UNIX timestamp
+// ARGV[4] -> stats expiration timestamp
+var retryCmd = redis.NewScript(`
+redis.call("LREM", KEYS[1], 0, ARGV[1])
+redis.call("ZADD", KEYS[2], ARGV[3], ARGV[2])
+local n = redis.call("INCR", KEYS[3])
+if tonumber(n) == 1 then
+	redis.call("EXPIREAT", KEYS[3], ARGV[4])
+end
+local m = redis.call("INCR", KEYS[4])
+if tonumber(m) == 1 then
+	redis.call("EXPIREAT", KEYS[4], ARGV[4])
+end
+return redis.status_reply("OK")`)
+
 // Retry moves the task from in-progress to retry queue, incrementing retry count
 // and assigning error message to the task message.
 func (r *RDB) Retry(msg *base.TaskMessage, processAt time.Time, errMsg string) error {
@@ -185,32 +211,11 @@ func (r *RDB) Retry(msg *base.TaskMessage, processAt time.Time, errMsg string) e
 	if err != nil {
 		return err
 	}
-	// KEYS[1] -> asynq:in_progress
-	// KEYS[2] -> asynq:retry
-	// KEYS[3] -> asynq:processed:<yyyy-mm-dd>
-	// KEYS[4] -> asynq:failure:<yyyy-mm-dd>
-	// ARGV[1] -> base.TaskMessage value to remove from base.InProgressQueue queue
-	// ARGV[2] -> base.TaskMessage value to add to Retry queue
-	// ARGV[3] -> retry_at UNIX timestamp
-	// ARGV[4] -> stats expiration timestamp
-	script := redis.NewScript(`
-	redis.call("LREM", KEYS[1], 0, ARGV[1])
-	redis.call("ZADD", KEYS[2], ARGV[3], ARGV[2])
-	local n = redis.call("INCR", KEYS[3])
-	if tonumber(n) == 1 then
-		redis.call("EXPIREAT", KEYS[3], ARGV[4])
-	end
-	local m = redis.call("INCR", KEYS[4])
-	if tonumber(m) == 1 then
-		redis.call("EXPIREAT", KEYS[4], ARGV[4])
-	end
-	return redis.status_reply("OK")
-	`)
 	now := time.Now()
 	processedKey := base.ProcessedKey(now)
 	failureKey := base.FailureKey(now)
 	expireAt := now.Add(statsTTL)
-	return script.Run(r.client,
+	return retryCmd.Run(r.client,
 		[]string{base.InProgressQueue, base.RetryQueue, processedKey, failureKey},
 		string(bytesToRemove), string(bytesToAdd), processAt.Unix(), expireAt.Unix()).Err()
 }
@@ -220,6 +225,31 @@ const (
 	deadExpirationInDays = 90
 )

+// KEYS[1] -> asynq:in_progress
+// KEYS[2] -> asynq:dead
+// KEYS[3] -> asynq:processed:<yyyy-mm-dd>
+// KEYS[4] -> asynq.failure:<yyyy-mm-dd>
+// ARGV[1] -> base.TaskMessage value to remove from base.InProgressQueue queue
+// ARGV[2] -> base.TaskMessage value to add to Dead queue
+// ARGV[3] -> died_at UNIX timestamp
+// ARGV[4] -> cutoff timestamp (e.g., 90 days ago)
+// ARGV[5] -> max number of tasks in dead queue (e.g., 100)
+// ARGV[6] -> stats expiration timestamp
+var killCmd = redis.NewScript(`
+redis.call("LREM", KEYS[1], 0, ARGV[1])
+redis.call("ZADD", KEYS[2], ARGV[3], ARGV[2])
+redis.call("ZREMRANGEBYSCORE", KEYS[2], "-inf", ARGV[4])
+redis.call("ZREMRANGEBYRANK", KEYS[2], 0, -ARGV[5])
+local n = redis.call("INCR", KEYS[3])
+if tonumber(n) == 1 then
+	redis.call("EXPIREAT", KEYS[3], ARGV[6])
+end
+local m = redis.call("INCR", KEYS[4])
+if tonumber(m) == 1 then
+	redis.call("EXPIREAT", KEYS[4], ARGV[6])
+end
+return redis.status_reply("OK")`)
+
 // Kill sends the task to "dead" queue from in-progress queue, assigning
 // the error message to the task.
 // It also trims the set by timestamp and set size.
@@ -239,47 +269,27 @@ func (r *RDB) Kill(msg *base.TaskMessage, errMsg string) error {
 	processedKey := base.ProcessedKey(now)
 	failureKey := base.FailureKey(now)
 	expireAt := now.Add(statsTTL)
-	// KEYS[1] -> asynq:in_progress
-	// KEYS[2] -> asynq:dead
-	// KEYS[3] -> asynq:processed:<yyyy-mm-dd>
-	// KEYS[4] -> asynq.failure:<yyyy-mm-dd>
-	// ARGV[1] -> base.TaskMessage value to remove from base.InProgressQueue queue
-	// ARGV[2] -> base.TaskMessage value to add to Dead queue
-	// ARGV[3] -> died_at UNIX timestamp
-	// ARGV[4] -> cutoff timestamp (e.g., 90 days ago)
-	// ARGV[5] -> max number of tasks in dead queue (e.g., 100)
-	// ARGV[6] -> stats expiration timestamp
-	script := redis.NewScript(`
-	redis.call("LREM", KEYS[1], 0, ARGV[1])
-	redis.call("ZADD", KEYS[2], ARGV[3], ARGV[2])
-	redis.call("ZREMRANGEBYSCORE", KEYS[2], "-inf", ARGV[4])
-	redis.call("ZREMRANGEBYRANK", KEYS[2], 0, -ARGV[5])
-	local n = redis.call("INCR", KEYS[3])
-	if tonumber(n) == 1 then
-		redis.call("EXPIREAT", KEYS[3], ARGV[6])
-	end
-	local m = redis.call("INCR", KEYS[4])
-	if tonumber(m) == 1 then
-		redis.call("EXPIREAT", KEYS[4], ARGV[6])
-	end
-	return redis.status_reply("OK")
-	`)
-	return script.Run(r.client,
+	return killCmd.Run(r.client,
 		[]string{base.InProgressQueue, base.DeadQueue, processedKey, failureKey},
 		string(bytesToRemove), string(bytesToAdd), now.Unix(), limit, maxDeadTasks, expireAt.Unix()).Err()
 }

-// RestoreUnfinished  moves all tasks from in-progress list to the queue
+// KEYS[1] -> asynq:in_progress
+// ARGV[1] -> queue prefix
+var requeueAllCmd = redis.NewScript(`
+local msgs = redis.call("LRANGE", KEYS[1], 0, -1)
+for _, msg in ipairs(msgs) do
+	local decoded = cjson.decode(msg)
+	local qkey = ARGV[1] .. decoded["Queue"]
+	redis.call("RPUSH", qkey, msg)
+	redis.call("LREM", KEYS[1], 0, msg)
+end
+return table.getn(msgs)`)
+
+// RequeueAll moves all tasks from in-progress list to the queue
 // and reports the number of tasks restored.
-func (r *RDB) RestoreUnfinished() (int64, error) {
-	script := redis.NewScript(`
-	local len = redis.call("LLEN", KEYS[1])
-	for i = len, 1, -1 do
-		redis.call("RPOPLPUSH", KEYS[1], KEYS[2])
-	end
-	return len
-	`)
-	res, err := script.Run(r.client, []string{base.InProgressQueue, base.DefaultQueue}).Result()
+func (r *RDB) RequeueAll() (int64, error) {
+	res, err := requeueAllCmd.Run(r.client, []string{base.InProgressQueue}, base.QueuePrefix).Result()
 	if err != nil {
 		return 0, err
 	}
@@ -310,36 +320,109 @@ func (r *RDB) CheckAndEnqueue(qnames ...string) error {
 	return nil
 }

+// KEYS[1] -> source queue (e.g. scheduled or retry queue)
+// ARGV[1] -> current unix time
+// ARGV[2] -> queue prefix
+var forwardCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], "-inf", ARGV[1])
+for _, msg in ipairs(msgs) do
+	local decoded = cjson.decode(msg)
+	local qkey = ARGV[2] .. decoded["Queue"]
+	redis.call("LPUSH", qkey, msg)
+	redis.call("ZREM", KEYS[1], msg)
+end
+return msgs`)
+
 // forward moves all tasks with a score less than the current unix time
 // from the src zset.
 func (r *RDB) forward(src string) error {
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], "-inf", ARGV[1])
-	for _, msg in ipairs(msgs) do
-		redis.call("ZREM", KEYS[1], msg)
-		local decoded = cjson.decode(msg)
-		local qkey = ARGV[2] .. decoded["Queue"]
-		redis.call("LPUSH", qkey, msg)
-	end
-	return msgs
-	`)
 	now := float64(time.Now().Unix())
-	return script.Run(r.client,
+	return forwardCmd.Run(r.client,
 		[]string{src}, now, base.QueuePrefix).Err()
 }

+// KEYS[1] -> source queue (e.g. scheduled or retry queue)
+// KEYS[2] -> destination queue
+var forwardSingleCmd = redis.NewScript(`
+local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], "-inf", ARGV[1])
+for _, msg in ipairs(msgs) do
+	redis.call("LPUSH", KEYS[2], msg)
+	redis.call("ZREM", KEYS[1], msg)
+end
+return msgs`)
+
 // forwardSingle moves all tasks with a score less than the current unix time
 // from the src zset to dst list.
 func (r *RDB) forwardSingle(src, dst string) error {
-	script := redis.NewScript(`
-	local msgs = redis.call("ZRANGEBYSCORE", KEYS[1], "-inf", ARGV[1])
-	for _, msg in ipairs(msgs) do
-		redis.call("ZREM", KEYS[1], msg)
-		redis.call("LPUSH", KEYS[2], msg)
-	end
-	return msgs
-	`)
 	now := float64(time.Now().Unix())
-	return script.Run(r.client,
+	return forwardSingleCmd.Run(r.client,
 		[]string{src, dst}, now).Err()
 }
+
+// KEYS[1] -> asynq:ps
+// KEYS[2] -> asynq:ps:<host:pid>
+// ARGV[1] -> expiration time
+// ARGV[2] -> TTL in seconds
+// ARGV[3] -> process info
+var writeProcessInfoCmd = redis.NewScript(`
+redis.call("ZADD", KEYS[1], ARGV[1], KEYS[2])
+redis.call("SETEX", KEYS[2], ARGV[2], ARGV[3])
+return redis.status_reply("OK")`)
+
+// WriteProcessInfo writes process information to redis with expiration
+// set to the value ttl.
+func (r *RDB) WriteProcessInfo(ps *base.ProcessInfo, ttl time.Duration) error {
+	bytes, err := json.Marshal(ps)
+	if err != nil {
+		return err
+	}
+	// Note: Add key to ZSET with expiration time as score.
+	// ref: https://github.com/antirez/redis/issues/135#issuecomment-2361996
+	exp := time.Now().Add(ttl).UTC()
+	key := base.ProcessInfoKey(ps.Host, ps.PID)
+	return writeProcessInfoCmd.Run(r.client, []string{base.AllProcesses, key}, float64(exp.Unix()), ttl.Seconds(), string(bytes)).Err()
+}
+
+// ReadProcessInfo reads process information stored in redis.
+func (r *RDB) ReadProcessInfo(host string, pid int) (*base.ProcessInfo, error) {
+	key := base.ProcessInfoKey(host, pid)
+	data, err := r.client.Get(key).Result()
+	if err != nil {
+		return nil, err
+	}
+	var pinfo base.ProcessInfo
+	err = json.Unmarshal([]byte(data), &pinfo)
+	if err != nil {
+		return nil, err
+	}
+	return &pinfo, nil
+}
+
+// KEYS[1] -> asynq:ps
+// KEYS[2] -> asynq:ps:<host:pid>
+var clearProcessInfoCmd = redis.NewScript(`
+redis.call("ZREM", KEYS[1], KEYS[2])
+redis.call("DEL", KEYS[2])
+return redis.status_reply("OK")`)
+
+// ClearProcessInfo deletes process information from redis.
+func (r *RDB) ClearProcessInfo(ps *base.ProcessInfo) error {
+	key := base.ProcessInfoKey(ps.Host, ps.PID)
+	return clearProcessInfoCmd.Run(r.client, []string{base.AllProcesses, key}).Err()
+}
+
+// CancelationPubSub returns a pubsub for cancelation messages.
+func (r *RDB) CancelationPubSub() (*redis.PubSub, error) {
+	pubsub := r.client.Subscribe(base.CancelChannel)
+	_, err := pubsub.Receive()
+	if err != nil {
+		return nil, err
+	}
+	return pubsub, nil
+}
+
+// PublishCancelation publish cancelation message to all subscribers.
+// The message is the ID for the task to be canceled.
+func (r *RDB) PublishCancelation(id string) error {
+	return r.client.Publish(base.CancelChannel, id).Err()
+}
--- a/internal/rdb/rdb_test.go
+++ b/internal/rdb/rdb_test.go
@@ -6,11 +6,13 @@ package rdb

 import (
 	"fmt"
+	"strconv"
 	"testing"
 	"time"

 	"github.com/go-redis/redis/v7"
 	"github.com/google/go-cmp/cmp"
+	"github.com/google/go-cmp/cmp/cmpopts"
 	h "github.com/hibiken/asynq/internal/asynqtest"
 	"github.com/hibiken/asynq/internal/base"
 )
@@ -236,33 +238,57 @@ func TestRequeue(t *testing.T) {
 	r := setup(t)
 	t1 := h.NewTaskMessage("send_email", nil)
 	t2 := h.NewTaskMessage("export_csv", nil)
+	t3 := h.NewTaskMessageWithQueue("send_email", nil, "critical")

 	tests := []struct {
-		enqueued       []*base.TaskMessage // initial state of the default queue
-		inProgress     []*base.TaskMessage // initial state of the in-progress list
-		target         *base.TaskMessage   // task to requeue
-		wantEnqueued   []*base.TaskMessage // final state of the default queue
-		wantInProgress []*base.TaskMessage // final state of the in-progress list
+		enqueued       map[string][]*base.TaskMessage // initial state of queues
+		inProgress     []*base.TaskMessage            // initial state of the in-progress list
+		target         *base.TaskMessage              // task to requeue
+		wantEnqueued   map[string][]*base.TaskMessage // final state of queues
+		wantInProgress []*base.TaskMessage            // final state of the in-progress list
 	}{
 		{
-			enqueued:       []*base.TaskMessage{},
-			inProgress:     []*base.TaskMessage{t1, t2},
-			target:         t1,
-			wantEnqueued:   []*base.TaskMessage{t1},
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {},
+			},
+			inProgress: []*base.TaskMessage{t1, t2},
+			target:     t1,
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1},
+			},
 			wantInProgress: []*base.TaskMessage{t2},
 		},
 		{
-			enqueued:       []*base.TaskMessage{t1},
-			inProgress:     []*base.TaskMessage{t2},
-			target:         t2,
-			wantEnqueued:   []*base.TaskMessage{t1, t2},
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1},
+			},
+			inProgress: []*base.TaskMessage{t2},
+			target:     t2,
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1, t2},
+			},
 			wantInProgress: []*base.TaskMessage{},
 		},
+		{
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1},
+				"critical":            {},
+			},
+			inProgress: []*base.TaskMessage{t2, t3},
+			target:     t3,
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1},
+				"critical":            {t3},
+			},
+			wantInProgress: []*base.TaskMessage{t2},
+		},
 	}

 	for _, tc := range tests {
 		h.FlushDB(t, r.client) // clean up db before each test case
-		h.SeedEnqueuedQueue(t, r.client, tc.enqueued)
+		for qname, msgs := range tc.enqueued {
+			h.SeedEnqueuedQueue(t, r.client, msgs, qname)
+		}
 		h.SeedInProgressQueue(t, r.client, tc.inProgress)

 		err := r.Requeue(tc.target)
@@ -271,9 +297,11 @@ func TestRequeue(t *testing.T) {
 			continue
 		}

-		gotEnqueued := h.GetEnqueuedMessages(t, r.client)
-		if diff := cmp.Diff(tc.wantEnqueued, gotEnqueued, h.SortMsgOpt); diff != "" {
-			t.Errorf("mismatch found in %q: (-want, +got):\n%s", base.DefaultQueue, diff)
+		for qname, want := range tc.wantEnqueued {
+			gotEnqueued := h.GetEnqueuedMessages(t, r.client, qname)
+			if diff := cmp.Diff(want, gotEnqueued, h.SortMsgOpt); diff != "" {
+				t.Errorf("mismatch found in %q; (-want, +got)\n%s", base.QueueKey(qname), diff)
+			}
 		}

 		gotInProgress := h.GetInProgressMessages(t, r.client)
@@ -513,50 +541,81 @@ func TestKill(t *testing.T) {
 	}
 }

-func TestRestoreUnfinished(t *testing.T) {
+func TestRequeueAll(t *testing.T) {
 	r := setup(t)
 	t1 := h.NewTaskMessage("send_email", nil)
 	t2 := h.NewTaskMessage("export_csv", nil)
 	t3 := h.NewTaskMessage("sync_stuff", nil)
+	t4 := h.NewTaskMessageWithQueue("important", nil, "critical")
+	t5 := h.NewTaskMessageWithQueue("minor", nil, "low")

 	tests := []struct {
 		inProgress     []*base.TaskMessage
-		enqueued       []*base.TaskMessage
+		enqueued       map[string][]*base.TaskMessage
 		want           int64
 		wantInProgress []*base.TaskMessage
-		wantEnqueued   []*base.TaskMessage
+		wantEnqueued   map[string][]*base.TaskMessage
 	}{
 		{
-			inProgress:     []*base.TaskMessage{t1, t2, t3},
-			enqueued:       []*base.TaskMessage{},
+			inProgress: []*base.TaskMessage{t1, t2, t3},
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {},
+			},
 			want:           3,
 			wantInProgress: []*base.TaskMessage{},
-			wantEnqueued:   []*base.TaskMessage{t1, t2, t3},
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1, t2, t3},
+			},
 		},
 		{
-			inProgress:     []*base.TaskMessage{},
-			enqueued:       []*base.TaskMessage{t1, t2, t3},
+			inProgress: []*base.TaskMessage{},
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1, t2, t3},
+			},
 			want:           0,
 			wantInProgress: []*base.TaskMessage{},
-			wantEnqueued:   []*base.TaskMessage{t1, t2, t3},
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1, t2, t3},
+			},
 		},
 		{
-			inProgress:     []*base.TaskMessage{t2, t3},
-			enqueued:       []*base.TaskMessage{t1},
+			inProgress: []*base.TaskMessage{t2, t3},
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1},
+			},
 			want:           2,
 			wantInProgress: []*base.TaskMessage{},
-			wantEnqueued:   []*base.TaskMessage{t1, t2, t3},
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1, t2, t3},
+			},
+		},
+		{
+			inProgress: []*base.TaskMessage{t2, t3, t4, t5},
+			enqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1},
+				"critical":            {},
+				"low":                 {},
+			},
+			want:           4,
+			wantInProgress: []*base.TaskMessage{},
+			wantEnqueued: map[string][]*base.TaskMessage{
+				base.DefaultQueueName: {t1, t2, t3},
+				"critical":            {t4},
+				"low":                 {t5},
+			},
 		},
 	}

 	for _, tc := range tests {
 		h.FlushDB(t, r.client) // clean up db before each test case
 		h.SeedInProgressQueue(t, r.client, tc.inProgress)
-		h.SeedEnqueuedQueue(t, r.client, tc.enqueued)
+		for qname, msgs := range tc.enqueued {
+			h.SeedEnqueuedQueue(t, r.client, msgs, qname)
+		}

-		got, err := r.RestoreUnfinished()
+		got, err := r.RequeueAll()
 		if got != tc.want || err != nil {
-			t.Errorf("(*RDB).RestoreUnfinished() = %v %v, want %v nil", got, err, tc.want)
+			t.Errorf("(*RDB).RequeueAll() = %v %v, want %v nil", got, err, tc.want)
 			continue
 		}

@@ -565,9 +624,11 @@ func TestRestoreUnfinished(t *testing.T) {
 			t.Errorf("mismatch found in %q: (-want, +got):\n%s", base.InProgressQueue, diff)
 		}

-		gotEnqueued := h.GetEnqueuedMessages(t, r.client)
-		if diff := cmp.Diff(tc.wantEnqueued, gotEnqueued, h.SortMsgOpt); diff != "" {
-			t.Errorf("mismatch found in %q: (-want, +got):\n%s", base.DefaultQueue, diff)
+		for qname, want := range tc.wantEnqueued {
+			gotEnqueued := h.GetEnqueuedMessages(t, r.client, qname)
+			if diff := cmp.Diff(want, gotEnqueued, h.SortMsgOpt); diff != "" {
+				t.Errorf("mismatch found in %q: (-want, +got):\n%s", base.QueueKey(qname), diff)
+			}
 		}
 	}
 }
@@ -679,3 +740,82 @@ func TestCheckAndEnqueue(t *testing.T) {
 		}
 	}
 }
+
+func TestReadWriteClearProcessInfo(t *testing.T) {
+	r := setup(t)
+	pinfo := &base.ProcessInfo{
+		Concurrency:       10,
+		Queues:            map[string]int{"default": 2, "email": 5, "low": 1},
+		PID:               98765,
+		Host:              "localhost",
+		State:             "running",
+		Started:           time.Now(),
+		ActiveWorkerCount: 1,
+	}
+
+	tests := []struct {
+		pi  *base.ProcessInfo
+		ttl time.Duration
+	}{
+		{pinfo, 5 * time.Second},
+	}
+
+	for _, tc := range tests {
+		h.FlushDB(t, r.client)
+
+		err := r.WriteProcessInfo(tc.pi, tc.ttl)
+		if err != nil {
+			t.Errorf("r.WriteProcessInfo returned an error: %v", err)
+			continue
+		}
+
+		got, err := r.ReadProcessInfo(tc.pi.Host, tc.pi.PID)
+		if err != nil {
+			t.Errorf("r.ReadProcessInfo returned an error: %v", err)
+			continue
+		}
+
+		ignoreOpt := cmpopts.IgnoreUnexported(base.ProcessInfo{})
+		if diff := cmp.Diff(tc.pi, got, ignoreOpt); diff != "" {
+			t.Errorf("r.ReadProcessInfo(%q, %d) = %+v, want %+v; (-want,+got)\n%s",
+				tc.pi.Host, tc.pi.PID, got, tc.pi, diff)
+		}
+
+		key := base.ProcessInfoKey(tc.pi.Host, tc.pi.PID)
+		gotTTL := r.client.TTL(key).Val()
+		if !cmp.Equal(tc.ttl, gotTTL, timeCmpOpt) {
+			t.Errorf("redis TTL %q returned %v, want %v", key, gotTTL, tc.ttl)
+		}
+
+		now := time.Now().UTC()
+		allKeys, err := r.client.ZRangeByScore(base.AllProcesses, &redis.ZRangeBy{
+			Min: strconv.Itoa(int(now.Unix())),
+			Max: "+inf",
+		}).Result()
+		if err != nil {
+			t.Errorf("redis ZRANGEBYSCORE %q %d +inf returned an error: %v",
+				base.AllProcesses, now.Unix(), err)
+			continue
+		}
+
+		wantAllKeys := []string{key}
+		if diff := cmp.Diff(wantAllKeys, allKeys); diff != "" {
+			t.Errorf("all keys = %v, want %v; (-want,+got)\n%s", allKeys, wantAllKeys, diff)
+		}
+
+		if err := r.ClearProcessInfo(tc.pi); err != nil {
+			t.Errorf("r.ClearProcessInfo returned an error: %v", err)
+			continue
+		}
+
+		// 1 means key exists
+		if r.client.Exists(key).Val() == 1 {
+			t.Errorf("expected %q to be deleted", key)
+		}
+
+		if r.client.ZCard(base.AllProcesses).Val() != 0 {
+			t.Errorf("expected %q to be empty", base.AllProcesses)
+		}
+
+	}
+}
--- a/processor.go
+++ b/processor.go
@@ -5,6 +5,7 @@
 package asynq

 import (
+	"context"
 	"fmt"
 	"math/rand"
 	"sort"
@@ -19,9 +20,11 @@ import (
 type processor struct {
 	rdb *rdb.RDB

+	pinfo *base.ProcessInfo
+
 	handler Handler

-	queueConfig map[string]uint
+	queueConfig map[string]int

 	// orderedQueues is set only in strict-priority mode.
 	orderedQueues []string
@@ -48,34 +51,34 @@ type processor struct {

 	// quit channel communicates to the in-flight worker goroutines to stop.
 	quit chan struct{}
+
+	// cancelations is a set of cancel functions for all in-progress tasks.
+	cancelations *base.Cancelations
 }

 type retryDelayFunc func(n int, err error, task *Task) time.Duration

 // newProcessor constructs a new processor.
-//
-// r is an instance of RDB used by the processor.
-// n specifies the max number of concurrenct worker goroutines.
-// qfcg is a mapping of queue names to associated priority level.
-// strict specifies whether queue priority should be treated strictly.
-// fn is a function to compute retry delay.
-func newProcessor(r *rdb.RDB, n int, qcfg map[string]uint, strict bool, fn retryDelayFunc, syncRequestCh chan<- *syncRequest) *processor {
+func newProcessor(r *rdb.RDB, pinfo *base.ProcessInfo, fn retryDelayFunc, syncRequestCh chan<- *syncRequest, cancelations *base.Cancelations) *processor {
+	qcfg := normalizeQueueCfg(pinfo.Queues)
 	orderedQueues := []string(nil)
-	if strict {
+	if pinfo.StrictPriority {
 		orderedQueues = sortByPriority(qcfg)
 	}
 	return &processor{
 		rdb:            r,
+		pinfo:          pinfo,
 		queueConfig:    qcfg,
 		orderedQueues:  orderedQueues,
 		retryDelayFunc: fn,
 		syncRequestCh:  syncRequestCh,
+		cancelations:   cancelations,
 		errLogLimiter:  rate.NewLimiter(rate.Every(3*time.Second), 1),
-		sema:           make(chan struct{}, n),
+		sema:           make(chan struct{}, pinfo.Concurrency),
 		done:           make(chan struct{}),
 		abort:          make(chan struct{}),
 		quit:           make(chan struct{}),
-		handler:        HandlerFunc(func(t *Task) error { return fmt.Errorf("handler not set") }),
+		handler:        HandlerFunc(func(ctx context.Context, t *Task) error { return fmt.Errorf("handler not set") }),
 	}
 }

@@ -100,6 +103,12 @@ func (p *processor) terminate() {
 	const timeout = 8 * time.Second
 	time.AfterFunc(timeout, func() { close(p.quit) })
 	logger.info("Waiting for all workers to finish...")
+
+	// send cancellation signal to all in-progress task handlers
+	for _, cancel := range p.cancelations.GetAll() {
+		cancel()
+	}
+
 	// block until all workers have released the token
 	for i := 0; i < cap(p.sema); i++ {
 		p.sema <- struct{}{}
@@ -153,13 +162,20 @@ func (p *processor) exec() {
 		p.requeue(msg)
 		return
 	case p.sema <- struct{}{}: // acquire token
+		p.pinfo.IncrActiveWorkerCount(1)
 		go func() {
-			defer func() { <-p.sema /* release token */ }()
+			defer func() {
+				<-p.sema /* release token */
+				p.pinfo.IncrActiveWorkerCount(-1)
+			}()

 			resCh := make(chan error, 1)
 			task := NewTask(msg.Type, msg.Payload)
+			ctx, cancel := createContext(msg)
+			p.cancelations.Add(msg.ID.String(), cancel)
 			go func() {
-				resCh <- perform(p.handler, task)
+				resCh <- perform(ctx, task, p.handler)
+				p.cancelations.Delete(msg.ID.String())
 			}()

 			select {
@@ -189,7 +205,7 @@ func (p *processor) exec() {
 // restore moves all tasks from "in-progress" back to queue
 // to restore all unfinished tasks.
 func (p *processor) restore() {
-	n, err := p.rdb.RestoreUnfinished()
+	n, err := p.rdb.RequeueAll()
 	if err != nil {
 		logger.error("Could not restore unfinished tasks: %v", err)
 	}
@@ -280,13 +296,13 @@ func (p *processor) queues() []string {
 // perform calls the handler with the given task.
 // If the call returns without panic, it simply returns the value,
 // otherwise, it recovers from panic and returns an error.
-func perform(h Handler, task *Task) (err error) {
+func perform(ctx context.Context, task *Task, h Handler) (err error) {
 	defer func() {
 		if x := recover(); x != nil {
 			err = fmt.Errorf("panic: %v", x)
 		}
 	}()
-	return h.ProcessTask(task)
+	return h.ProcessTask(ctx, task)
 }

 // uniq dedupes elements and returns a slice of unique names of length l.
@@ -308,7 +324,7 @@ func uniq(names []string, l int) []string {

 // sortByPriority returns a list of queue names sorted by
 // their priority level in descending order.
-func sortByPriority(qcfg map[string]uint) []string {
+func sortByPriority(qcfg map[string]int) []string {
 	var queues []*queue
 	for qname, n := range qcfg {
 		queues = append(queues, &queue{qname, n})
@@ -323,7 +339,7 @@ func sortByPriority(qcfg map[string]uint) []string {

 type queue struct {
 	name     string
-	priority uint
+	priority int
 }

 type byPriority []*queue
@@ -331,3 +347,48 @@ type byPriority []*queue
 func (x byPriority) Len() int           { return len(x) }
 func (x byPriority) Less(i, j int) bool { return x[i].priority < x[j].priority }
 func (x byPriority) Swap(i, j int)      { x[i], x[j] = x[j], x[i] }
+
+// normalizeQueueCfg divides priority numbers by their
+// greatest common divisor.
+func normalizeQueueCfg(queueCfg map[string]int) map[string]int {
+	var xs []int
+	for _, x := range queueCfg {
+		xs = append(xs, x)
+	}
+	d := gcd(xs...)
+	res := make(map[string]int)
+	for q, x := range queueCfg {
+		res[q] = x / d
+	}
+	return res
+}
+
+func gcd(xs ...int) int {
+	fn := func(x, y int) int {
+		for y > 0 {
+			x, y = y, x%y
+		}
+		return x
+	}
+	res := xs[0]
+	for i := 0; i < len(xs); i++ {
+		res = fn(xs[i], res)
+		if res == 1 {
+			return 1
+		}
+	}
+	return res
+}
+
+// createContext returns a context and cancel function for a given task message.
+func createContext(msg *base.TaskMessage) (context.Context, context.CancelFunc) {
+	timeout, err := time.ParseDuration(msg.Timeout)
+	if err != nil {
+		logger.error("cannot parse timeout duration for %+v", msg)
+		return context.WithCancel(context.Background())
+	}
+	if timeout == 0 {
+		return context.WithCancel(context.Background())
+	}
+	return context.WithTimeout(context.Background(), timeout)
+}
--- a/processor_test.go
+++ b/processor_test.go
@@ -5,6 +5,7 @@
 package asynq

 import (
+	"context"
 	"fmt"
 	"sort"
 	"sync"
@@ -59,13 +60,15 @@ func TestProcessorSuccess(t *testing.T) {
 		// instantiate a new processor
 		var mu sync.Mutex
 		var processed []*Task
-		handler := func(task *Task) error {
+		handler := func(ctx context.Context, task *Task) error {
 			mu.Lock()
 			defer mu.Unlock()
 			processed = append(processed, task)
 			return nil
 		}
-		p := newProcessor(rdbClient, 10, defaultQueueConfig, false, defaultDelayFunc, nil)
+		pi := base.NewProcessInfo("localhost", 1234, 10, defaultQueueConfig, false)
+		cancelations := base.NewCancelations()
+		p := newProcessor(rdbClient, pi, defaultDelayFunc, nil, cancelations)
 		p.handler = HandlerFunc(handler)

 		p.start()
@@ -145,10 +148,12 @@ func TestProcessorRetry(t *testing.T) {
 		delayFunc := func(n int, e error, t *Task) time.Duration {
 			return tc.delay
 		}
-		handler := func(task *Task) error {
+		handler := func(ctx context.Context, task *Task) error {
 			return fmt.Errorf(errMsg)
 		}
-		p := newProcessor(rdbClient, 10, defaultQueueConfig, false, delayFunc, nil)
+		pi := base.NewProcessInfo("localhost", 1234, 10, defaultQueueConfig, false)
+		cancelations := base.NewCancelations()
+		p := newProcessor(rdbClient, pi, delayFunc, nil, cancelations)
 		p.handler = HandlerFunc(handler)

 		p.start()
@@ -187,11 +192,11 @@ func TestProcessorQueues(t *testing.T) {
 	})

 	tests := []struct {
-		queueCfg map[string]uint
+		queueCfg map[string]int
 		want     []string
 	}{
 		{
-			queueCfg: map[string]uint{
+			queueCfg: map[string]int{
 				"high":    6,
 				"default": 3,
 				"low":     1,
@@ -199,7 +204,7 @@ func TestProcessorQueues(t *testing.T) {
 			want: []string{"high", "default", "low"},
 		},
 		{
-			queueCfg: map[string]uint{
+			queueCfg: map[string]int{
 				"default": 1,
 			},
 			want: []string{"default"},
@@ -207,7 +212,9 @@ func TestProcessorQueues(t *testing.T) {
 	}

 	for _, tc := range tests {
-		p := newProcessor(nil, 10, tc.queueCfg, false, defaultDelayFunc, nil)
+		pi := base.NewProcessInfo("localhost", 1234, 10, tc.queueCfg, false)
+		cancelations := base.NewCancelations()
+		p := newProcessor(nil, pi, defaultDelayFunc, nil, cancelations)
 		got := p.queues()
 		if diff := cmp.Diff(tc.want, got, sortOpt); diff != "" {
 			t.Errorf("with queue config: %v\n(*processor).queues() = %v, want %v\n(-want,+got):\n%s",
@@ -261,19 +268,21 @@ func TestProcessorWithStrictPriority(t *testing.T) {
 		// instantiate a new processor
 		var mu sync.Mutex
 		var processed []*Task
-		handler := func(task *Task) error {
+		handler := func(ctx context.Context, task *Task) error {
 			mu.Lock()
 			defer mu.Unlock()
 			processed = append(processed, task)
 			return nil
 		}
-		queueCfg := map[string]uint{
+		queueCfg := map[string]int{
 			"critical":            3,
 			base.DefaultQueueName: 2,
 			"low":                 1,
 		}
 		// Note: Set concurrency to 1 to make sure tasks are processed one at a time.
-		p := newProcessor(rdbClient, 1 /*concurrency */, queueCfg, true /* strict */, defaultDelayFunc, nil)
+		pi := base.NewProcessInfo("localhost", 1234, 1 /*concurrency */, queueCfg, true /* strict */)
+		cancelations := base.NewCancelations()
+		p := newProcessor(rdbClient, pi, defaultDelayFunc, nil, cancelations)
 		p.handler = HandlerFunc(handler)

 		p.start()
@@ -299,7 +308,7 @@ func TestPerform(t *testing.T) {
 	}{
 		{
 			desc: "handler returns nil",
-			handler: func(t *Task) error {
+			handler: func(ctx context.Context, t *Task) error {
 				return nil
 			},
 			task:    NewTask("gen_thumbnail", map[string]interface{}{"src": "some/img/path"}),
@@ -307,7 +316,7 @@ func TestPerform(t *testing.T) {
 		},
 		{
 			desc: "handler returns error",
-			handler: func(t *Task) error {
+			handler: func(ctx context.Context, t *Task) error {
 				return fmt.Errorf("something went wrong")
 			},
 			task:    NewTask("gen_thumbnail", map[string]interface{}{"src": "some/img/path"}),
@@ -315,7 +324,7 @@ func TestPerform(t *testing.T) {
 		},
 		{
 			desc: "handler panics",
-			handler: func(t *Task) error {
+			handler: func(ctx context.Context, t *Task) error {
 				panic("something went terribly wrong")
 			},
 			task:    NewTask("gen_thumbnail", map[string]interface{}{"src": "some/img/path"}),
@@ -324,7 +333,7 @@ func TestPerform(t *testing.T) {
 	}

 	for _, tc := range tests {
-		got := perform(tc.handler, tc.task)
+		got := perform(context.Background(), tc.task, tc.handler)
 		if !tc.wantErr && got != nil {
 			t.Errorf("%s: perform() = %v, want nil", tc.desc, got)
 			continue
--- a/scheduler.go
+++ b/scheduler.go
@@ -23,7 +23,7 @@ type scheduler struct {
 	qnames []string
 }

-func newScheduler(r *rdb.RDB, avgInterval time.Duration, qcfg map[string]uint) *scheduler {
+func newScheduler(r *rdb.RDB, avgInterval time.Duration, qcfg map[string]int) *scheduler {
 	var qnames []string
 	for q := range qcfg {
 		qnames = append(qnames, q)
--- a/subscriber.go
+++ b/subscriber.go
@@ -0,0 +1,58 @@
+// 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 (
+	"github.com/hibiken/asynq/internal/base"
+	"github.com/hibiken/asynq/internal/rdb"
+)
+
+type subscriber struct {
+	rdb *rdb.RDB
+
+	// channel to communicate back to the long running "subscriber" goroutine.
+	done chan struct{}
+
+	// cancelations hold cancel functions for all in-progress tasks.
+	cancelations *base.Cancelations
+}
+
+func newSubscriber(rdb *rdb.RDB, cancelations *base.Cancelations) *subscriber {
+	return &subscriber{
+		rdb:          rdb,
+		done:         make(chan struct{}),
+		cancelations: cancelations,
+	}
+}
+
+func (s *subscriber) terminate() {
+	logger.info("Subscriber shutting down...")
+	// Signal the subscriber goroutine to stop.
+	s.done <- struct{}{}
+}
+
+func (s *subscriber) start() {
+	pubsub, err := s.rdb.CancelationPubSub()
+	cancelCh := pubsub.Channel()
+	if err != nil {
+		logger.error("cannot subscribe to cancelation channel: %v", err)
+		return
+	}
+	go func() {
+		for {
+			select {
+			case <-s.done:
+				pubsub.Close()
+				logger.info("Subscriber done")
+				return
+			case msg := <-cancelCh:
+				cancel := s.cancelations.Get(msg.Payload)
+				if cancel != nil {
+					cancel()
+				}
+			}
+		}
+	}()
+}
--- a/subscriber_test.go
+++ b/subscriber_test.go
@@ -0,0 +1,57 @@
+// 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 (
+	"testing"
+	"time"
+
+	"github.com/hibiken/asynq/internal/base"
+	"github.com/hibiken/asynq/internal/rdb"
+)
+
+func TestSubscriber(t *testing.T) {
+	r := setup(t)
+	rdbClient := rdb.NewRDB(r)
+
+	tests := []struct {
+		registeredID string // ID for which cancel func is registered
+		publishID    string // ID to be published
+		wantCalled   bool   // whether cancel func should be called
+	}{
+		{"abc123", "abc123", true},
+		{"abc456", "abc123", false},
+	}
+
+	for _, tc := range tests {
+		called := false
+		fakeCancelFunc := func() {
+			called = true
+		}
+		cancelations := base.NewCancelations()
+		cancelations.Add(tc.registeredID, fakeCancelFunc)
+
+		subscriber := newSubscriber(rdbClient, cancelations)
+		subscriber.start()
+
+		if err := rdbClient.PublishCancelation(tc.publishID); err != nil {
+			subscriber.terminate()
+			t.Fatalf("could not publish cancelation message: %v", err)
+		}
+
+		// allow for redis to publish message
+		time.Sleep(time.Second)
+
+		if called != tc.wantCalled {
+			if tc.wantCalled {
+				t.Errorf("fakeCancelFunc was not called, want the function to be called")
+			} else {
+				t.Errorf("fakeCancelFunc was called, want the function to not be called")
+			}
+		}
+
+		subscriber.terminate()
+	}
+}
--- a/syncer_test.go
+++ b/syncer_test.go
@@ -87,6 +87,7 @@ func TestSyncerRetry(t *testing.T) {
 		t.Errorf("%q has length %d; want %d", base.InProgressQueue, l, len(inProgress))
 	}

+	// FIXME: This assignment introduces data race and running the test with -race will fail.
 	// simualate failover.
 	rdbClient = rdb.NewRDB(goodClient)

--- a/tools/asynqmon/README.md
+++ b/tools/asynqmon/README.md
@@ -1,6 +1,6 @@
 # Asynqmon

-Asynqmon is a CLI tool to monitor the queues managed by `asynq` package.
+Asynqmon is a command line tool to monitor the tasks managed by `asynq` package.

 ## Table of Contents

@@ -8,10 +8,12 @@ Asynqmon is a CLI tool to monitor the queues managed by `asynq` package.
 - [Quick Start](#quick-start)
  - [Stats](#stats)
  - [History](#history)
+  - [Process Status](#process-status)
  - [List](#list)
  - [Enqueue](#enqueue)
  - [Delete](#delete)
  - [Kill](#kill)
+  - [Cancel](#cancel)
 - [Config File](#config-file)

 ## Installation
@@ -22,9 +24,9 @@ In order to use the tool, compile it using the following command:

 This will create the asynqmon executable under your `$GOPATH/bin` directory.

-## Quick Start
+## Quickstart

-Asynqmon tool has a few commands to inspect the state of tasks and queues.
+The tool has a few commands to inspect the state of tasks and queues.

 Run `asynqmon help` to see all the available commands.

@@ -34,7 +36,7 @@ By default, Asynqmon will try to connect to a redis server running at `localhost

 ### Stats

-Stats command gives the overview of the current state of tasks and queues. Run it in conjunction with `watch` command to repeatedly run `stats`.
+Stats command gives the overview of the current state of tasks and queues. You can run it in conjunction with `watch` command to repeatedly run `stats`.

 Example:

@@ -46,35 +48,115 @@ This will run `asynqmon stats` command every 3 seconds.

 ### History

-TODO: Add discription
+History command shows the number of processed and failed tasks from the last x days.
+
+By default, it shows the stats from the last 10 days. Use `--days` to specify the number of days.
+
+Example:
+
+    asynqmon history --days=30
+
+![Gif](/docs/assets/asynqmon_history.gif)
+
+### Process Status
+
+PS (ProcessStatus) command shows the list of running worker processes.
+
+Example:
+
+    asynqmon ps
+
+![Gif](/docs/assets/asynqmon_ps.gif)

 ### List

-TODO: Add discription
+List command shows all tasks in the specified state in a table format
+
+Example:
+
+    asynqmon ls retry
+    asynqmon ls scheduled
+    asynqmon ls dead
+    asynqmon ls enqueued:default
+    asynqmon ls inprogress

 ### Enqueue

-TODO: Add discription
+There are two commands to enqueue tasks.
+
+Command `enq` takes a task ID and moves the task to **Enqueued** state. You can obtain the task ID by running `ls` command.
+
+Example:
+
+    asynqmon enq d:1575732274:bnogo8gt6toe23vhef0g
+
+Command `enqall` moves all tasks to **Enqueued** state from the specified state.
+
+Example:
+
+    asynqmon enqall retry
+
+Running the above command will move all **Retry** tasks to **Enqueued** state.

 ### Delete

-TODO: Add discription
+There are two commands for task deletion.
+
+Command `del` takes a task ID and deletes the task. You can obtain the task ID by running `ls` command.
+
+Example:
+
+    asynqmon del r:1575732274:bnogo8gt6toe23vhef0g
+
+Command `delall` deletes all tasks which are in the specified state.
+
+Example:
+
+    asynqmon delall retry
+
+Running the above command will delete all **Retry** tasks.

 ### Kill

-TODO: Add discription
+There are two commands to kill (i.e. move to dead state) tasks.
+
+Command `kill` takes a task ID and kills the task. You can obtain the task ID by running `ls` command.
+
+Example:
+
+    asynqmon kill r:1575732274:bnogo8gt6toe23vhef0g
+
+Command `killall` kills all tasks which are in the specified state.
+
+Example:
+
+    asynqmon killall retry
+
+Running the above command will move all **Retry** tasks to **Dead** state.
+
+### Cancel
+
+Command `cancel` takes a task ID and sends a cancelation signal to the goroutine processing the specified task.  
+You can obtain the task ID by running `ls` command.
+
+The task should be in "in-progress" state.
+Handler implementation needs to be context aware in order to actually stop processing.
+
+Example:
+
+    asynqmon cancel bnogo8gt6toe23vhef0g

 ## Config File

-You can use a config file to set default values for flags.
+You can use a config file to set default values for the flags.
 This is useful, for example when you have to connect to a remote redis server.

 By default, `asynqmon` will try to read config file located in
-`$HOME/.asynqmon.(yml|json)`. You can specify the file location via `--config` flag.
+`$HOME/.asynqmon.(yaml|json)`. You can specify the file location via `--config` flag.

 Config file example:

-```yml
+```yaml
 uri: 127.0.0.1:6379
 db: 2
 password: mypassword
--- a/tools/asynqmon/cmd/cancel.go
+++ b/tools/asynqmon/cmd/cancel.go
@@ -0,0 +1,53 @@
+// 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 cmd
+
+import (
+	"fmt"
+	"os"
+
+	"github.com/go-redis/redis/v7"
+	"github.com/hibiken/asynq/internal/rdb"
+	"github.com/spf13/cobra"
+	"github.com/spf13/viper"
+)
+
+// cancelCmd represents the cancel command
+var cancelCmd = &cobra.Command{
+	Use:   "cancel [task id]",
+	Short: "Sends a cancelation signal to the goroutine processing the specified task",
+	Long: `Cancel (asynqmon cancel) will send a cancelation signal to the goroutine processing 
+the specified task. 
+
+The command takes one argument which specifies the task to cancel.
+The task should be in in-progress state.
+Identifier for a task should be obtained by running "asynqmon ls" command.
+
+Handler implementation needs to be context aware for cancelation signal to
+actually cancel the processing.
+
+Example: asynqmon cancel bnogo8gt6toe23vhef0g`,
+	Args: cobra.ExactArgs(1),
+	Run:  cancel,
+}
+
+func init() {
+	rootCmd.AddCommand(cancelCmd)
+}
+
+func cancel(cmd *cobra.Command, args []string) {
+	r := rdb.NewRDB(redis.NewClient(&redis.Options{
+		Addr:     viper.GetString("uri"),
+		DB:       viper.GetInt("db"),
+		Password: viper.GetString("password"),
+	}))
+
+	err := r.PublishCancelation(args[0])
+	if err != nil {
+		fmt.Printf("could not send cancelation signal: %v\n", err)
+		os.Exit(1)
+	}
+	fmt.Printf("Successfully sent cancelation siganl for task %s\n", args[0])
+}
--- a/tools/asynqmon/cmd/delall.go
+++ b/tools/asynqmon/cmd/delall.go
@@ -19,7 +19,7 @@ var delallValidArgs = []string{"scheduled", "retry", "dead"}
 // delallCmd represents the delall command
 var delallCmd = &cobra.Command{
 	Use:   "delall [state]",
-	Short: "Deletes all tasks from the specified state",
+	Short: "Deletes all tasks in the specified state",
 	Long: `Delall (asynqmon delall) will delete all tasks in the specified state.

 The argument should be one of "scheduled", "retry", or "dead".
--- a/tools/asynqmon/cmd/killall.go
+++ b/tools/asynqmon/cmd/killall.go
@@ -19,7 +19,7 @@ var killallValidArgs = []string{"scheduled", "retry"}
 // killallCmd represents the killall command
 var killallCmd = &cobra.Command{
 	Use:   "killall [state]",
-	Short: "Update all tasks to dead state from the specified state",
+	Short: "Kills all tasks in the specified state",
 	Long: `Killall (asynqmon killall) will update all tasks from the specified state to dead state.

 The argument should be either "scheduled" or "retry".
--- a/tools/asynqmon/cmd/ls.go
+++ b/tools/asynqmon/cmd/ls.go
@@ -10,7 +10,6 @@ import (
 	"os"
 	"strconv"
 	"strings"
-	"text/tabwriter"
 	"time"

 	"github.com/go-redis/redis/v7"
@@ -35,29 +34,34 @@ The argument value should be one of "enqueued", "inprogress", "scheduled",
 Example:
 asynqmon ls dead -> Lists all tasks in dead state

-Enqueued tasks can optionally be filtered by providing queue names after ":"
+Enqueued tasks requires a queue name after ":"
 Example:
-asynqmon ls enqueued:critical -> List tasks from critical queue only
+asynqmon ls enqueued:default  -> List tasks from default queue
+asynqmon ls enqueued:critical -> List tasks from critical queue 
 `,
 	Args: cobra.ExactValidArgs(1),
 	Run:  ls,
 }

+// Flags
+var pageSize int
+var pageNum int
+
 func init() {
 	rootCmd.AddCommand(lsCmd)
-
-	// Here you will define your flags and configuration settings.
-
-	// Cobra supports Persistent Flags which will work for this command
-	// and all subcommands, e.g.:
-	// lsCmd.PersistentFlags().String("foo", "", "A help for foo")
-
-	// Cobra supports local flags which will only run when this command
-	// is called directly, e.g.:
-	// lsCmd.Flags().BoolP("toggle", "t", false, "Help message for toggle")
+	lsCmd.Flags().IntVar(&pageSize, "size", 30, "page size")
+	lsCmd.Flags().IntVar(&pageNum, "page", 0, "page number - zero indexed (default 0)")
 }

 func ls(cmd *cobra.Command, args []string) {
+	if pageSize < 0 {
+		fmt.Println("page size cannot be negative.")
+		os.Exit(1)
+	}
+	if pageNum < 0 {
+		fmt.Println("page number cannot be negative.")
+		os.Exit(1)
+	}
 	c := redis.NewClient(&redis.Options{
 		Addr:     viper.GetString("uri"),
 		DB:       viper.GetInt("db"),
@@ -67,7 +71,11 @@ func ls(cmd *cobra.Command, args []string) {
 	parts := strings.Split(args[0], ":")
 	switch parts[0] {
 	case "enqueued":
-		listEnqueued(r, parts[1:]...)
+		if len(parts) != 2 {
+			fmt.Printf("error: Missing queue name\n`asynqmon ls enqueued:[queue name]`\n")
+			os.Exit(1)
+		}
+		listEnqueued(r, parts[1])
 	case "inprogress":
 		listInProgress(r)
 	case "scheduled":
@@ -77,7 +85,7 @@ func ls(cmd *cobra.Command, args []string) {
 	case "dead":
 		listDead(r)
 	default:
-		fmt.Printf("error: `asynqmon ls [state]` only accepts %v as the argument.\n", lsValidArgs)
+		fmt.Printf("error: `asynqmon ls [state]`\nonly accepts %v as the argument.\n", lsValidArgs)
 		os.Exit(1)
 	}
 }
@@ -113,24 +121,14 @@ func parseQueryID(queryID string) (id xid.ID, score int64, qtype string, err err
 	return id, score, qtype, nil
 }

-func listEnqueued(r *rdb.RDB, qnames ...string) {
-	tasks, err := r.ListEnqueued(qnames...)
+func listEnqueued(r *rdb.RDB, qname string) {
+	tasks, err := r.ListEnqueued(qname, rdb.Pagination{Size: pageSize, Page: pageNum})
 	if err != nil {
 		fmt.Println(err)
 		os.Exit(1)
 	}
 	if len(tasks) == 0 {
-		msg := "No enqueued tasks"
-		if len(qnames) > 0 {
-			msg += " in"
-			for i, q := range qnames {
-				msg += fmt.Sprintf(" %q queue", q)
-				if i != len(qnames)-1 {
-					msg += ","
-				}
-			}
-		}
-		fmt.Println(msg)
+		fmt.Printf("No enqueued tasks in %q queue\n", qname)
 		return
 	}
 	cols := []string{"ID", "Type", "Payload", "Queue"}
@@ -140,10 +138,11 @@ func listEnqueued(r *rdb.RDB, qnames ...string) {
 		}
 	}
 	printTable(cols, printRows)
+	fmt.Printf("\nShowing %d tasks from page %d\n", len(tasks), pageNum)
 }

 func listInProgress(r *rdb.RDB) {
-	tasks, err := r.ListInProgress()
+	tasks, err := r.ListInProgress(rdb.Pagination{Size: pageSize, Page: pageNum})
 	if err != nil {
 		fmt.Println(err)
 		os.Exit(1)
@@ -159,10 +158,11 @@ func listInProgress(r *rdb.RDB) {
 		}
 	}
 	printTable(cols, printRows)
+	fmt.Printf("\nShowing %d tasks from page %d\n", len(tasks), pageNum)
 }

 func listScheduled(r *rdb.RDB) {
-	tasks, err := r.ListScheduled()
+	tasks, err := r.ListScheduled(rdb.Pagination{Size: pageSize, Page: pageNum})
 	if err != nil {
 		fmt.Println(err)
 		os.Exit(1)
@@ -179,10 +179,11 @@ func listScheduled(r *rdb.RDB) {
 		}
 	}
 	printTable(cols, printRows)
+	fmt.Printf("\nShowing %d tasks from page %d\n", len(tasks), pageNum)
 }

 func listRetry(r *rdb.RDB) {
-	tasks, err := r.ListRetry()
+	tasks, err := r.ListRetry(rdb.Pagination{Size: pageSize, Page: pageNum})
 	if err != nil {
 		fmt.Println(err)
 		os.Exit(1)
@@ -191,18 +192,24 @@ func listRetry(r *rdb.RDB) {
 		fmt.Println("No retry tasks")
 		return
 	}
-	cols := []string{"ID", "Type", "Payload", "Retry In", "Last Error", "Retried", "Max Retry", "Queue"}
+	cols := []string{"ID", "Type", "Payload", "Next Retry", "Last Error", "Retried", "Max Retry", "Queue"}
 	printRows := func(w io.Writer, tmpl string) {
 		for _, t := range tasks {
-			retryIn := fmt.Sprintf("%.0f seconds", t.ProcessAt.Sub(time.Now()).Seconds())
-			fmt.Fprintf(w, tmpl, queryID(t.ID, t.Score, "r"), t.Type, t.Payload, retryIn, t.ErrorMsg, t.Retried, t.Retry, t.Queue)
+			var nextRetry string
+			if d := t.ProcessAt.Sub(time.Now()); d > 0 {
+				nextRetry = fmt.Sprintf("in %v", d.Round(time.Second))
+			} else {
+				nextRetry = "right now"
+			}
+			fmt.Fprintf(w, tmpl, queryID(t.ID, t.Score, "r"), t.Type, t.Payload, nextRetry, t.ErrorMsg, t.Retried, t.Retry, t.Queue)
 		}
 	}
 	printTable(cols, printRows)
+	fmt.Printf("\nShowing %d tasks from page %d\n", len(tasks), pageNum)
 }

 func listDead(r *rdb.RDB) {
-	tasks, err := r.ListDead()
+	tasks, err := r.ListDead(rdb.Pagination{Size: pageSize, Page: pageNum})
 	if err != nil {
 		fmt.Println(err)
 		os.Exit(1)
@@ -218,19 +225,5 @@ func listDead(r *rdb.RDB) {
 		}
 	}
 	printTable(cols, printRows)
-}
-
-func printTable(cols []string, printRows func(w io.Writer, tmpl string)) {
-	format := strings.Repeat("%v\t", len(cols)) + "\n"
-	tw := new(tabwriter.Writer).Init(os.Stdout, 0, 8, 2, ' ', 0)
-	var headers []interface{}
-	var seps []interface{}
-	for _, name := range cols {
-		headers = append(headers, name)
-		seps = append(seps, strings.Repeat("-", len(name)))
-	}
-	fmt.Fprintf(tw, format, headers...)
-	fmt.Fprintf(tw, format, seps...)
-	printRows(tw, format)
-	tw.Flush()
+	fmt.Printf("\nShowing %d tasks from page %d\n", len(tasks), pageNum)
 }
--- a/tools/asynqmon/cmd/ps.go
+++ b/tools/asynqmon/cmd/ps.go
@@ -0,0 +1,118 @@
+// 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 cmd
+
+import (
+	"fmt"
+	"io"
+	"os"
+	"sort"
+	"strings"
+	"time"
+
+	"github.com/go-redis/redis/v7"
+	"github.com/hibiken/asynq/internal/rdb"
+	"github.com/spf13/cobra"
+	"github.com/spf13/viper"
+)
+
+// psCmd represents the ps command
+var psCmd = &cobra.Command{
+	Use:   "ps",
+	Short: "Shows all background worker processes",
+	Long: `Ps (asynqmon ps) will show all background worker processes
+backed by the specified redis instance.
+
+The command shows the following for each process:
+* Host and PID of the process
+* Number of active workers out of worker pool
+* Queue configuration
+* State of the worker process ("running" | "stopped")
+* Time the process was started
+
+A "running" process is processing tasks in queues.
+A "stopped" process is no longer processing new tasks.`,
+	Args: cobra.NoArgs,
+	Run:  ps,
+}
+
+func init() {
+	rootCmd.AddCommand(psCmd)
+}
+
+func ps(cmd *cobra.Command, args []string) {
+	r := rdb.NewRDB(redis.NewClient(&redis.Options{
+		Addr:     viper.GetString("uri"),
+		DB:       viper.GetInt("db"),
+		Password: viper.GetString("password"),
+	}))
+
+	processes, err := r.ListProcesses()
+	if err != nil {
+		fmt.Println(err)
+		os.Exit(1)
+	}
+	if len(processes) == 0 {
+		fmt.Println("No processes")
+		return
+	}
+
+	// sort by hostname and pid
+	sort.Slice(processes, func(i, j int) bool {
+		x, y := processes[i], processes[j]
+		if x.Host != y.Host {
+			return x.Host < y.Host
+		}
+		return x.PID < y.PID
+	})
+
+	// print processes
+	cols := []string{"Host", "PID", "State", "Active Workers", "Queues", "Started"}
+	printRows := func(w io.Writer, tmpl string) {
+		for _, ps := range processes {
+			fmt.Fprintf(w, tmpl,
+				ps.Host, ps.PID, ps.State,
+				fmt.Sprintf("%d/%d", ps.ActiveWorkerCount, ps.Concurrency),
+				formatQueues(ps.Queues), timeAgo(ps.Started))
+		}
+	}
+	printTable(cols, printRows)
+}
+
+// timeAgo takes a time and returns a string of the format "<duration> ago".
+func timeAgo(since time.Time) string {
+	d := time.Since(since).Round(time.Second)
+	return fmt.Sprintf("%v ago", d)
+}
+
+func formatQueues(qmap map[string]int) string {
+	// sort queues by priority and name
+	type queue struct {
+		name     string
+		priority int
+	}
+	var queues []*queue
+	for qname, p := range qmap {
+		queues = append(queues, &queue{qname, p})
+	}
+	sort.Slice(queues, func(i, j int) bool {
+		x, y := queues[i], queues[j]
+		if x.priority != y.priority {
+			return x.priority > y.priority
+		}
+		return x.name < y.name
+	})
+
+	var b strings.Builder
+	l := len(queues)
+	for _, q := range queues {
+		fmt.Fprintf(&b, "%s:%d", q.name, q.priority)
+		l--
+		if l > 0 {
+			b.WriteString(" ")
+		}
+	}
+	return b.String()
+}
--- a/tools/asynqmon/cmd/root.go
+++ b/tools/asynqmon/cmd/root.go
@@ -6,7 +6,10 @@ package cmd

 import (
 	"fmt"
+	"io"
 	"os"
+	"strings"
+	"text/tabwriter"

 	"github.com/spf13/cobra"

@@ -25,14 +28,7 @@ var password string
 var rootCmd = &cobra.Command{
 	Use:   "asynqmon",
 	Short: "A monitoring tool for asynq queues",
-	Long: `Asynqmon is a CLI tool to inspect tasks and queues managed by asynq package.
-
-Use commands to query and mutate the current state of tasks and queues.
-
-Monitoring commands such as "stats" and "ls" can be used in conjunction with the
-"watch" command to continuously run the command at a certain interval.
-	
-Example: watch -n 5 asynqmon stats`,
+	Long:  `Asynqmon is a montoring CLI to inspect tasks and queues managed by asynq.`,
 }

 // Execute adds all child commands to the root command and sets flags appropriately.
@@ -81,3 +77,36 @@ func initConfig() {
 		fmt.Println("Using config file:", viper.ConfigFileUsed())
 	}
 }
+
+// printTable is a helper function to print data in table format.
+//
+// cols is a list of headers and printRow specifies how to print rows.
+//
+// Example:
+// type User struct {
+//     Name string
+//     Addr string
+//     Age  int
+// }
+// data := []*User{{"user1", "addr1", 24}, {"user2", "addr2", 42}, ...}
+// cols := []string{"Name", "Addr", "Age"}
+// printRows := func(w io.Writer, tmpl string) {
+//     for _, u := range data {
+//         fmt.Fprintf(w, tmpl, u.Name, u.Addr, u.Age)
+//     }
+// }
+// printTable(cols, printRows)
+func printTable(cols []string, printRows func(w io.Writer, tmpl string)) {
+	format := strings.Repeat("%v\t", len(cols)) + "\n"
+	tw := new(tabwriter.Writer).Init(os.Stdout, 0, 8, 2, ' ', 0)
+	var headers []interface{}
+	var seps []interface{}
+	for _, name := range cols {
+		headers = append(headers, name)
+		seps = append(seps, strings.Repeat("-", len(name)))
+	}
+	fmt.Fprintf(tw, format, headers...)
+	fmt.Fprintf(tw, format, seps...)
+	printRows(tw, format)
+	tw.Flush()
+}
Author	SHA1	Message	Date
Ken Hibino	0856ef32e0	v0.4.0	2020-02-13 06:25:27 -08:00
Ken Hibino	70efb7b455	[ci skip] Update changelog	2020-02-13 06:23:51 -08:00
Ken Hibino	d33ca98648	Use int as priority value type. Do not use unsigned int for merely non-negative quantities as it complicates simple arithmetic.	2020-02-13 06:23:51 -08:00
Ken Hibino	bf31fcc3ec	[ci skip] Update readme and changelog	2020-02-13 06:23:51 -08:00
Ken Hibino	cd55a0e316	Add cancel command to asynqmon cli	2020-02-13 06:23:51 -08:00
Ken Hibino	6685827147	Add redis pubsub subscriber for cancelation	2020-02-13 06:23:51 -08:00
Ken Hibino	133bb6c2c6	Update docs for context and timeout	2020-02-13 06:23:51 -08:00
Ken Hibino	39459b4412	Allow user to specify timeout per task	2020-02-13 06:23:51 -08:00
Ken Hibino	6e14062325	Send cancellation signal to all in-progress handlers on shutdown	2020-02-13 06:23:51 -08:00
Ken Hibino	1b41c721b4	Change Handler interface to take context.Context	2020-02-13 06:23:51 -08:00
Ken Hibino	79a04e52a3	[cip skip] Add link to intro video in readme	2020-02-09 10:59:32 -08:00
Ken Hibino	d9327cf24b	Avoid creating script struct on every rdb method invocation	2020-02-09 09:34:41 -08:00
Ken Hibino	7e33c9ebe0	[ci skip] Add contribution doc	2020-02-08 09:48:04 -08:00
Ken Hibino	a6b79eb7a0	[ci skip] Update readme	2020-02-07 06:45:36 -08:00
Ken Hibino	7af69c8d3c	[ci skip] Fix comments	2020-02-06 20:46:59 -08:00
Ken Hibino	6dd4c688f4	[ci skip] Update readme	2020-02-05 22:12:26 -08:00
Ken Hibino	024cfa9fe6	Update output text for task's next retry value	2020-02-05 22:12:26 -08:00
Ken Hibino	461a6f2587	v0.3.0	2020-02-04 06:29:57 -08:00
Ken Hibino	ba9b42f428	Clean up cli description text	2020-02-04 06:16:29 -08:00
Ken Hibino	3744072e9b	Sort processes by host and pid in ps output	2020-02-03 06:59:02 -08:00
Ken Hibino	eadfd5f8b4	[ci skip] Update changelog	2020-02-02 20:59:00 -08:00
Ken Hibino	96f06ac89b	Add ps command to asynqmon	2020-02-02 20:56:33 -08:00
Ken Hibino	d03fa34eaf	Add hearbeater	2020-02-01 09:35:49 -08:00
Ken Hibino	489e695433	[ci skip] Change font used in demo.gif	2020-01-30 21:47:08 -08:00
Ken Hibino	9ae4be8184	Run and compare benchmarks after successful ci-build	2020-01-30 21:38:16 -08:00
Ken Hibino	36af486303	[ci skip] Add prettier gif for demo	2020-01-29 21:36:17 -08:00
Ken Hibino	a46eb97e6f	[ci skip] Shorten readme	2020-01-26 20:06:52 -08:00
Ken Hibino	c934ef115b	v0.2.2	2020-01-26 16:07:44 -08:00
Ken Hibino	6fbaa2ed6c	(fix): RestoreUnfinished to select correct queue	2020-01-26 16:05:46 -08:00
Ken Hibino	166497748b	(fix): Requeue to select correct queue	2020-01-26 16:05:46 -08:00
Ken Hibino	31123fd42a	Paginate tasks with asynqmon ls command Changes: * Added --page and --size flags to ls command * By default, the command will show first 30 tasks from the specified queue	2020-01-26 13:12:01 -08:00
Ken Hibino	3ed155b45b	[ci skip] Update readme	2020-01-25 08:08:13 -08:00
Ken Hibino	58d2ed94e7	[ci skip] Fix typo	2020-01-23 06:05:18 -08:00