1
0
Fork 0
mirror of https://codeberg.org/forgejo/forgejo.git synced 2024-11-25 08:59:31 -05:00
forgejo/modules/queue/workerqueue_test.go
wxiaoguang 6f9c278559
Rewrite queue (#24505)
# ⚠️ Breaking

Many deprecated queue config options are removed (actually, they should
have been removed in 1.18/1.19).

If you see the fatal message when starting Gitea: "Please update your
app.ini to remove deprecated config options", please follow the error
messages to remove these options from your app.ini.

Example:

```
2023/05/06 19:39:22 [E] Removed queue option: `[indexer].ISSUE_INDEXER_QUEUE_TYPE`. Use new options in `[queue.issue_indexer]`
2023/05/06 19:39:22 [E] Removed queue option: `[indexer].UPDATE_BUFFER_LEN`. Use new options in `[queue.issue_indexer]`
2023/05/06 19:39:22 [F] Please update your app.ini to remove deprecated config options
```

Many options in `[queue]` are are dropped, including:
`WRAP_IF_NECESSARY`, `MAX_ATTEMPTS`, `TIMEOUT`, `WORKERS`,
`BLOCK_TIMEOUT`, `BOOST_TIMEOUT`, `BOOST_WORKERS`, they can be removed
from app.ini.

# The problem

The old queue package has some legacy problems:

* complexity: I doubt few people could tell how it works.
* maintainability: Too many channels and mutex/cond are mixed together,
too many different structs/interfaces depends each other.
* stability: due to the complexity & maintainability, sometimes there
are strange bugs and difficult to debug, and some code doesn't have test
(indeed some code is difficult to test because a lot of things are mixed
together).
* general applicability: although it is called "queue", its behavior is
not a well-known queue.
* scalability: it doesn't seem easy to make it work with a cluster
without breaking its behaviors.

It came from some very old code to "avoid breaking", however, its
technical debt is too heavy now. It's a good time to introduce a better
"queue" package.

# The new queue package

It keeps using old config and concept as much as possible.

* It only contains two major kinds of concepts:
    * The "base queue": channel, levelqueue, redis
* They have the same abstraction, the same interface, and they are
tested by the same testing code.
* The "WokerPoolQueue", it uses the "base queue" to provide "worker
pool" function, calls the "handler" to process the data in the base
queue.
* The new code doesn't do "PushBack"
* Think about a queue with many workers, the "PushBack" can't guarantee
the order for re-queued unhandled items, so in new code it just does
"normal push"
* The new code doesn't do "pause/resume"
* The "pause/resume" was designed to handle some handler's failure: eg:
document indexer (elasticsearch) is down
* If a queue is paused for long time, either the producers blocks or the
new items are dropped.
* The new code doesn't do such "pause/resume" trick, it's not a common
queue's behavior and it doesn't help much.
* If there are unhandled items, the "push" function just blocks for a
few seconds and then re-queue them and retry.
* The new code doesn't do "worker booster"
* Gitea's queue's handlers are light functions, the cost is only the
go-routine, so it doesn't make sense to "boost" them.
* The new code only use "max worker number" to limit the concurrent
workers.
* The new "Push" never blocks forever
* Instead of creating more and more blocking goroutines, return an error
is more friendly to the server and to the end user.

There are more details in code comments: eg: the "Flush" problem, the
strange "code.index" hanging problem, the "immediate" queue problem.

Almost ready for review.

TODO:

* [x] add some necessary comments during review
* [x] add some more tests if necessary
* [x] update documents and config options
* [x] test max worker / active worker
* [x] re-run the CI tasks to see whether any test is flaky
* [x] improve the `handleOldLengthConfiguration` to provide more
friendly messages
* [x] fine tune default config values (eg: length?)

## Code coverage:

![image](https://user-images.githubusercontent.com/2114189/236620635-55576955-f95d-4810-b12f-879026a3afdf.png)
2023-05-08 19:49:59 +08:00

260 lines
7.8 KiB
Go

// Copyright 2023 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package queue
import (
"context"
"strconv"
"sync"
"testing"
"time"
"code.gitea.io/gitea/modules/setting"
"github.com/stretchr/testify/assert"
)
func runWorkerPoolQueue[T any](q *WorkerPoolQueue[T]) func() {
var stop func()
started := make(chan struct{})
stopped := make(chan struct{})
go func() {
q.Run(func(f func()) { stop = f; close(started) }, nil)
close(stopped)
}()
<-started
return func() {
stop()
<-stopped
}
}
func TestWorkerPoolQueueUnhandled(t *testing.T) {
oldUnhandledItemRequeueDuration := unhandledItemRequeueDuration.Load()
unhandledItemRequeueDuration.Store(0)
defer unhandledItemRequeueDuration.Store(oldUnhandledItemRequeueDuration)
mu := sync.Mutex{}
test := func(t *testing.T, queueSetting setting.QueueSettings) {
queueSetting.Length = 100
queueSetting.Type = "channel"
queueSetting.Datadir = t.TempDir() + "/test-queue"
m := map[int]int{}
// odds are handled once, evens are handled twice
handler := func(items ...int) (unhandled []int) {
testRecorder.Record("handle:%v", items)
for _, item := range items {
mu.Lock()
if item%2 == 0 && m[item] == 0 {
unhandled = append(unhandled, item)
}
m[item]++
mu.Unlock()
}
return unhandled
}
q, _ := NewWorkerPoolQueueBySetting("test-workpoolqueue", queueSetting, handler, false)
stop := runWorkerPoolQueue(q)
for i := 0; i < queueSetting.Length; i++ {
testRecorder.Record("push:%v", i)
assert.NoError(t, q.Push(i))
}
assert.NoError(t, q.FlushWithContext(context.Background(), 0))
stop()
ok := true
for i := 0; i < queueSetting.Length; i++ {
if i%2 == 0 {
ok = ok && assert.EqualValues(t, 2, m[i], "test %s: item %d", t.Name(), i)
} else {
ok = ok && assert.EqualValues(t, 1, m[i], "test %s: item %d", t.Name(), i)
}
}
if !ok {
t.Logf("m: %v", m)
t.Logf("records: %v", testRecorder.Records())
}
testRecorder.Reset()
}
runCount := 2 // we can run these tests even hundreds times to see its stability
t.Run("1/1", func(t *testing.T) {
for i := 0; i < runCount; i++ {
test(t, setting.QueueSettings{BatchLength: 1, MaxWorkers: 1})
}
})
t.Run("3/1", func(t *testing.T) {
for i := 0; i < runCount; i++ {
test(t, setting.QueueSettings{BatchLength: 3, MaxWorkers: 1})
}
})
t.Run("4/5", func(t *testing.T) {
for i := 0; i < runCount; i++ {
test(t, setting.QueueSettings{BatchLength: 4, MaxWorkers: 5})
}
})
}
func TestWorkerPoolQueuePersistence(t *testing.T) {
runCount := 2 // we can run these tests even hundreds times to see its stability
t.Run("1/1", func(t *testing.T) {
for i := 0; i < runCount; i++ {
testWorkerPoolQueuePersistence(t, setting.QueueSettings{BatchLength: 1, MaxWorkers: 1, Length: 100})
}
})
t.Run("3/1", func(t *testing.T) {
for i := 0; i < runCount; i++ {
testWorkerPoolQueuePersistence(t, setting.QueueSettings{BatchLength: 3, MaxWorkers: 1, Length: 100})
}
})
t.Run("4/5", func(t *testing.T) {
for i := 0; i < runCount; i++ {
testWorkerPoolQueuePersistence(t, setting.QueueSettings{BatchLength: 4, MaxWorkers: 5, Length: 100})
}
})
}
func testWorkerPoolQueuePersistence(t *testing.T, queueSetting setting.QueueSettings) {
testCount := queueSetting.Length
queueSetting.Type = "level"
queueSetting.Datadir = t.TempDir() + "/test-queue"
mu := sync.Mutex{}
var tasksQ1, tasksQ2 []string
q1 := func() {
startWhenAllReady := make(chan struct{}) // only start data consuming when the "testCount" tasks are all pushed into queue
stopAt20Shutdown := make(chan struct{}) // stop and shutdown at the 20th item
testHandler := func(data ...string) []string {
<-startWhenAllReady
time.Sleep(10 * time.Millisecond)
for _, s := range data {
mu.Lock()
tasksQ1 = append(tasksQ1, s)
mu.Unlock()
if s == "task-20" {
close(stopAt20Shutdown)
}
}
return nil
}
q, _ := NewWorkerPoolQueueBySetting("pr_patch_checker_test", queueSetting, testHandler, true)
stop := runWorkerPoolQueue(q)
for i := 0; i < testCount; i++ {
_ = q.Push("task-" + strconv.Itoa(i))
}
close(startWhenAllReady)
<-stopAt20Shutdown // it's possible to have more than 20 tasks executed
stop()
}
q1() // run some tasks and shutdown at an intermediate point
time.Sleep(100 * time.Millisecond) // because the handler in q1 has a slight delay, we need to wait for it to finish
q2 := func() {
testHandler := func(data ...string) []string {
for _, s := range data {
mu.Lock()
tasksQ2 = append(tasksQ2, s)
mu.Unlock()
}
return nil
}
q, _ := NewWorkerPoolQueueBySetting("pr_patch_checker_test", queueSetting, testHandler, true)
stop := runWorkerPoolQueue(q)
assert.NoError(t, q.FlushWithContext(context.Background(), 0))
stop()
}
q2() // restart the queue to continue to execute the tasks in it
assert.NotZero(t, len(tasksQ1))
assert.NotZero(t, len(tasksQ2))
assert.EqualValues(t, testCount, len(tasksQ1)+len(tasksQ2))
}
func TestWorkerPoolQueueActiveWorkers(t *testing.T) {
oldWorkerIdleDuration := workerIdleDuration
workerIdleDuration = 300 * time.Millisecond
defer func() {
workerIdleDuration = oldWorkerIdleDuration
}()
handler := func(items ...int) (unhandled []int) {
time.Sleep(100 * time.Millisecond)
return nil
}
q, _ := NewWorkerPoolQueueBySetting("test-workpoolqueue", setting.QueueSettings{Type: "channel", BatchLength: 1, MaxWorkers: 1, Length: 100}, handler, false)
stop := runWorkerPoolQueue(q)
for i := 0; i < 5; i++ {
assert.NoError(t, q.Push(i))
}
time.Sleep(50 * time.Millisecond)
assert.EqualValues(t, 1, q.GetWorkerNumber())
assert.EqualValues(t, 1, q.GetWorkerActiveNumber())
time.Sleep(500 * time.Millisecond)
assert.EqualValues(t, 1, q.GetWorkerNumber())
assert.EqualValues(t, 0, q.GetWorkerActiveNumber())
time.Sleep(workerIdleDuration)
assert.EqualValues(t, 1, q.GetWorkerNumber()) // there is at least one worker after the queue begins working
stop()
q, _ = NewWorkerPoolQueueBySetting("test-workpoolqueue", setting.QueueSettings{Type: "channel", BatchLength: 1, MaxWorkers: 3, Length: 100}, handler, false)
stop = runWorkerPoolQueue(q)
for i := 0; i < 15; i++ {
assert.NoError(t, q.Push(i))
}
time.Sleep(50 * time.Millisecond)
assert.EqualValues(t, 3, q.GetWorkerNumber())
assert.EqualValues(t, 3, q.GetWorkerActiveNumber())
time.Sleep(500 * time.Millisecond)
assert.EqualValues(t, 3, q.GetWorkerNumber())
assert.EqualValues(t, 0, q.GetWorkerActiveNumber())
time.Sleep(workerIdleDuration)
assert.EqualValues(t, 1, q.GetWorkerNumber()) // there is at least one worker after the queue begins working
stop()
}
func TestWorkerPoolQueueShutdown(t *testing.T) {
oldUnhandledItemRequeueDuration := unhandledItemRequeueDuration.Load()
unhandledItemRequeueDuration.Store(int64(100 * time.Millisecond))
defer unhandledItemRequeueDuration.Store(oldUnhandledItemRequeueDuration)
// simulate a slow handler, it doesn't handle any item (all items will be pushed back to the queue)
handlerCalled := make(chan struct{})
handler := func(items ...int) (unhandled []int) {
if items[0] == 0 {
close(handlerCalled)
}
time.Sleep(100 * time.Millisecond)
return items
}
qs := setting.QueueSettings{Type: "level", Datadir: t.TempDir() + "/queue", BatchLength: 3, MaxWorkers: 4, Length: 20}
q, _ := NewWorkerPoolQueueBySetting("test-workpoolqueue", qs, handler, false)
stop := runWorkerPoolQueue(q)
for i := 0; i < qs.Length; i++ {
assert.NoError(t, q.Push(i))
}
<-handlerCalled
time.Sleep(50 * time.Millisecond) // wait for a while to make sure all workers are active
assert.EqualValues(t, 4, q.GetWorkerActiveNumber())
stop() // stop triggers shutdown
assert.EqualValues(t, 0, q.GetWorkerActiveNumber())
// no item was ever handled, so we still get all of them again
q, _ = NewWorkerPoolQueueBySetting("test-workpoolqueue", qs, handler, false)
assert.EqualValues(t, 20, q.GetQueueItemNumber())
}