move implementation to internal/ directory (#828)

This commit is contained in:
Christian Schwarz
2024-10-18 19:21:17 +02:00
committed by GitHub
parent b9b9ad10cf
commit 908807bd59
360 changed files with 507 additions and 507 deletions
+422
View File
@@ -0,0 +1,422 @@
// package trace provides activity tracing via ctx through Tasks and Spans
//
// # Basic Concepts
//
// Tracing can be used to identify where a piece of code spends its time.
//
// The Go standard library provides package runtime/trace which is useful to identify CPU bottlenecks or
// to understand what happens inside the Go runtime.
// However, it is not ideal for application level tracing, in particular if those traces should be understandable
// to tech-savvy users (albeit not developers).
//
// This package provides the concept of Tasks and Spans to express what activity is happening within an application:
// - Neither task nor span is really tangible but instead contained within the context.Context tree
// - Tasks represent concurrent activity (i.e. goroutines).
// - Spans represent a semantic stack trace within a task.
// As a consequence, whenever a context is propagated across goroutine boundary, you need to create a child task:
//
// go func(ctx context.Context) {
// ctx, endTask = WithTask(ctx, "what-happens-inside-the-child-task")
// defer endTask()
// // ...
// }(ctx)
//
// Within the task, you can open up a hierarchy of spans.
// In contrast to tasks, which have can multiple concurrently running child tasks,
// spans must nest and not cross the goroutine boundary.
//
// ctx, endSpan = WithSpan(ctx, "copy-dir")
// defer endSpan()
// for _, f := range dir.Files() {
// func() {
// ctx, endSpan := WithSpan(ctx, fmt.Sprintf("copy-file %q", f))
// defer endspan()
// b, _ := ioutil.ReadFile(f)
// _ = ioutil.WriteFile(f + ".copy", b, 0600)
// }()
// }
//
// In combination:
//
// ctx, endTask = WithTask(ctx, "copy-dirs")
// defer endTask()
// for i := range dirs {
// go func(dir string) {
// ctx, endTask := WithTask(ctx, "copy-dir")
// defer endTask()
// for _, f := range filesIn(dir) {
// func() {
// ctx, endSpan := WithSpan(ctx, fmt.Sprintf("copy-file %q", f))
// defer endspan()
// b, _ := ioutil.ReadFile(f)
// _ = ioutil.WriteFile(f + ".copy", b, 0600)
// }()
// }
// }()
// }
//
// Note that a span ends at the time you call endSpan - not before and not after that.
// If you violate the stack-like nesting of spans by forgetting an endSpan() invocation,
// the out-of-order endSpan() will panic.
//
// A similar rule applies to the endTask closure returned by WithTask:
// If a task has live child tasks at the time you call endTask(), the call will panic.
//
// Recovering from endSpan() or endTask() panics will corrupt the trace stack and lead to corrupt tracefile output.
//
// # Best Practices For Naming Tasks And Spans
//
// Tasks should always have string constants as names, and must not contain the `#` character. WHy?
// First, the visualization by chrome://tracing draws a horizontal bar for each task in the trace.
// Also, the package appends `#NUM` for each concurrently running instance of a task name.
// Note that the `#NUM` suffix will be reused if a task has ended, in order to avoid an
// infinite number of horizontal bars in the visualization.
//
// # Chrome-compatible Tracefile Support
//
// The activity trace generated by usage of WithTask and WithSpan can be rendered to a JSON output file
// that can be loaded into chrome://tracing .
// Apart from function GetSpanStackOrDefault, this is the main benefit of this package.
//
// First, there is a convenience environment variable 'ZREPL_ACTIVITY_TRACE' that can be set to an output path.
// From process start onward, a trace is written to that path.
//
// More consumers can attach to the activity trace through the ChrometraceClientWebsocketHandler websocket handler.
//
// If a write error is encountered with any consumer (including the env-var based one), the consumer is closed and
// will not receive further trace output.
package trace
import (
"context"
"fmt"
"os"
runtimedebug "runtime/debug"
"strings"
"time"
"github.com/kr/pretty"
"github.com/pkg/errors"
"github.com/prometheus/client_golang/prometheus"
"github.com/zrepl/zrepl/internal/util/chainlock"
)
var metrics struct {
activeTasks prometheus.Gauge
}
var taskNamer *uniqueConcurrentTaskNamer = newUniqueTaskNamer()
func init() {
metrics.activeTasks = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: "zrepl",
Subsystem: "trace",
Name: "active_tasks",
Help: "number of active (tracing-level) tasks in the daemon",
})
}
func RegisterMetrics(r prometheus.Registerer) {
r.MustRegister(metrics.activeTasks)
}
type traceNode struct {
// NOTE: members with prefix debug are only valid if the package variable debugEnabled is true.
id string
annotation string
parentTask *traceNode
debugCreationStack string // debug: stack trace of when the traceNode was created
mtx chainlock.L
activeChildTasks int32 // only for task nodes, insignificant for span nodes
debugActiveChildTasks map[*traceNode]bool // debug: set of active child tasks
parentSpan *traceNode
activeChildSpan *traceNode // nil if task or span doesn't have an active child span
startedAt time.Time
endedAt time.Time
}
func (s *traceNode) StartedAt() time.Time { return s.startedAt }
func (s *traceNode) EndedAt() time.Time { return s.endedAt }
// caller must hold mtx
func (s *traceNode) debugString() string {
if !debugEnabled {
return "<debugEnabled=false>"
}
return pretty.Sprint(s)
}
// Returned from WithTask or WithSpan.
// Must be called once the task or span ends.
// See package-level docs for nesting rules.
// Wrong call order / forgetting to call it will result in panics.
type DoneFunc func()
var ErrTaskStillHasActiveChildTasks = fmt.Errorf("end task: task still has active child tasks")
// Start a new root task or create a child task of an existing task.
//
// This is required when starting a new goroutine and
// passing an existing task context to it.
//
// taskName should be a constantand must not contain '#'
//
// The implementation ensures that,
// if multiple tasks with the same name exist simultaneously,
// a unique suffix is appended to uniquely identify the task opened with this function.
func WithTask(ctx context.Context, taskName string) (context.Context, DoneFunc) {
var parentTask *traceNode
nodeI := ctx.Value(contextKeyTraceNode)
if nodeI != nil {
node := nodeI.(*traceNode)
if node.parentSpan != nil {
parentTask = node.parentTask
} else {
parentTask = node
}
}
// find the first ancestor that hasn't ended yet (nil if need be)
if parentTask != nil {
parentTask.mtx.Lock()
}
for parentTask != nil && !parentTask.endedAt.IsZero() {
thisParent := parentTask
parentTask = parentTask.parentTask
// lock parent first such that it isn't modified by other callers to thisParent
if parentTask != nil {
parentTask.mtx.Lock()
}
thisParent.mtx.Unlock()
}
// invariant: either parentTask != nil and we hold the lock on parentTask, or parentTask is nil
taskName, taskNameDone := taskNamer.UniqueConcurrentTaskName(taskName)
this := &traceNode{
id: genID(),
annotation: taskName,
parentTask: parentTask,
activeChildTasks: 0,
parentSpan: nil,
activeChildSpan: nil,
startedAt: time.Now(),
endedAt: time.Time{},
}
if debugEnabled {
this.debugCreationStack = string(runtimedebug.Stack())
this.debugActiveChildTasks = map[*traceNode]bool{}
}
if parentTask != nil {
this.parentTask.activeChildTasks++
if debugEnabled {
this.parentTask.debugActiveChildTasks[this] = true
}
parentTask.mtx.Unlock()
}
ctx = context.WithValue(ctx, contextKeyTraceNode, this)
chrometraceBeginTask(this)
metrics.activeTasks.Inc()
endTaskFunc := func() {
// only hold locks while manipulating the tree
// (trace writer might block too long and unlike spans, tasks are updated concurrently)
alreadyEnded := func() (alreadyEnded bool) {
if this.parentTask != nil {
defer this.parentTask.mtx.Lock().Unlock()
}
defer this.mtx.Lock().Unlock()
if this.activeChildTasks != 0 {
if debugEnabled {
// the debugString can be quite long and panic won't print it completely
fmt.Fprintf(os.Stderr, "going to panic due to activeChildTasks:\n%s\n", this.debugString())
}
panic(errors.WithMessagef(ErrTaskStillHasActiveChildTasks, "end task: %v active child tasks (run daemon with env var %s=1 for more details)\n", this.activeChildTasks, debugEnabledEnvVar))
}
// support idempotent task ends
if !this.endedAt.IsZero() {
return true
}
this.endedAt = time.Now()
if this.parentTask != nil {
this.parentTask.activeChildTasks--
if debugEnabled {
delete(this.parentTask.debugActiveChildTasks, this)
}
if this.parentTask.activeChildTasks < 0 {
if debugEnabled {
// the debugString can be quite long and panic won't print it completely
fmt.Fprintf(os.Stderr, "going to panic due to activeChildTasks < 0:\n%s\n", this.parentTask.debugString())
}
panic(fmt.Sprintf("impl error: parent task with negative activeChildTasks count: %v", this.parentTask.activeChildTasks))
}
}
return false
}()
if alreadyEnded {
return
}
chrometraceEndTask(this)
metrics.activeTasks.Dec()
taskNameDone()
}
return ctx, endTaskFunc
}
var ErrAlreadyActiveChildSpan = fmt.Errorf("create child span: span already has an active child span")
var ErrSpanStillHasActiveChildSpan = fmt.Errorf("end span: span still has active child spans")
// Start a new span.
// Important: ctx must have an active task (see WithTask)
func WithSpan(ctx context.Context, annotation string) (context.Context, DoneFunc) {
var parentSpan, parentTask *traceNode
nodeI := ctx.Value(contextKeyTraceNode)
if nodeI != nil {
parentSpan = nodeI.(*traceNode)
if parentSpan.parentSpan == nil {
parentTask = parentSpan
} else {
parentTask = parentSpan.parentTask
}
} else {
panic("must be called from within a task")
}
this := &traceNode{
id: genID(),
annotation: annotation,
parentTask: parentTask,
parentSpan: parentSpan,
activeChildSpan: nil,
startedAt: time.Now(),
endedAt: time.Time{},
}
if debugEnabled {
this.debugCreationStack = string(runtimedebug.Stack())
}
parentSpan.mtx.HoldWhile(func() {
if parentSpan.activeChildSpan != nil {
panic(ErrAlreadyActiveChildSpan)
}
parentSpan.activeChildSpan = this
})
ctx = context.WithValue(ctx, contextKeyTraceNode, this)
chrometraceBeginSpan(this)
callbackEndSpan := callbackBeginSpan(ctx)
endTaskFunc := func() {
defer parentSpan.mtx.Lock().Unlock()
if parentSpan.activeChildSpan != this && this.endedAt.IsZero() {
panic("impl error: activeChildSpan should not change while != nil because there can only be one")
}
defer this.mtx.Lock().Unlock()
if this.activeChildSpan != nil {
panic(ErrSpanStillHasActiveChildSpan)
}
if !this.endedAt.IsZero() {
return // support idempotent span ends
}
parentSpan.activeChildSpan = nil
this.endedAt = time.Now()
chrometraceEndSpan(this)
callbackEndSpan(this)
}
return ctx, endTaskFunc
}
type StackKind struct {
symbolizeTask func(t *traceNode) string
symbolizeSpan func(s *traceNode) string
}
var (
StackKindId = &StackKind{
symbolizeTask: func(t *traceNode) string { return t.id },
symbolizeSpan: func(s *traceNode) string { return s.id },
}
SpanStackKindCombined = &StackKind{
symbolizeTask: func(t *traceNode) string { return fmt.Sprintf("(%s %q)", t.id, t.annotation) },
symbolizeSpan: func(s *traceNode) string { return fmt.Sprintf("(%s %q)", s.id, s.annotation) },
}
SpanStackKindAnnotation = &StackKind{
symbolizeTask: func(t *traceNode) string { return t.annotation },
symbolizeSpan: func(s *traceNode) string { return s.annotation },
}
)
func (n *traceNode) task() *traceNode {
task := n.parentTask
if n.parentSpan == nil {
task = n
}
return task
}
func (n *traceNode) TaskName() string {
task := n.task()
return task.annotation
}
func (this *traceNode) TaskAndSpanStack(kind *StackKind) (spanIdStack string) {
task := this.task()
var spansInTask []*traceNode
for s := this; s != nil; s = s.parentSpan {
spansInTask = append(spansInTask, s)
}
var tasks []*traceNode
for t := task; t != nil; t = t.parentTask {
tasks = append(tasks, t)
}
var taskIdsRev []string
for i := len(tasks) - 1; i >= 0; i-- {
taskIdsRev = append(taskIdsRev, kind.symbolizeTask(tasks[i]))
}
var spanIdsRev []string
for i := len(spansInTask) - 1; i >= 0; i-- {
spanIdsRev = append(spanIdsRev, kind.symbolizeSpan(spansInTask[i]))
}
taskStack := strings.Join(taskIdsRev, "$")
return fmt.Sprintf("%s$%s", taskStack, strings.Join(spanIdsRev, "."))
}
func GetSpanStackOrDefault(ctx context.Context, kind StackKind, def string) string {
if nI := ctx.Value(contextKeyTraceNode); nI != nil {
n := nI.(*traceNode)
return n.TaskAndSpanStack(StackKindId)
} else {
return def
}
}
@@ -0,0 +1,54 @@
package trace
import (
"context"
"time"
"github.com/zrepl/zrepl/internal/util/chainlock"
)
type SpanInfo interface {
StartedAt() time.Time
EndedAt() time.Time
TaskAndSpanStack(kind *StackKind) string
}
type Callback struct {
OnBegin func(ctx context.Context)
OnEnd func(ctx context.Context, spanInfo SpanInfo)
}
var callbacks struct {
mtx chainlock.L
cs []Callback
}
func RegisterCallback(c Callback) {
callbacks.mtx.HoldWhile(func() {
callbacks.cs = append(callbacks.cs, c)
})
}
func callbackBeginSpan(ctx context.Context) func(SpanInfo) {
// capture the current state of callbacks into a local variable
// this is safe because the slice is append-only and immutable
// (it is important that a callback registered _after_ callbackBeginSpin is called does not get called on OnEnd)
var cbs []Callback
callbacks.mtx.HoldWhile(func() {
cbs = callbacks.cs
})
for _, cb := range cbs {
if cb.OnBegin != nil {
cb.OnBegin(ctx)
}
}
return func(spanInfo SpanInfo) {
for _, cb := range cbs {
if cb.OnEnd != nil {
cb.OnEnd(ctx, spanInfo)
}
}
}
}
@@ -0,0 +1,230 @@
package trace
// The functions in this file are concerned with the generation
// of trace files based on the information from WithTask and WithSpan.
//
// The emitted trace files are open-ended array of JSON objects
// that follow the Chrome trace file format:
// https://docs.google.com/document/d/1CvAClvFfyA5R-PhYUmn5OOQtYMH4h6I0nSsKchNAySU/preview
//
// The emitted JSON can be loaded into Chrome's chrome://tracing view.
//
// The trace file can be written to a file whose path is specified in an env file,
// and be written to web sockets established on ChrometraceHttpHandler
import (
"bufio"
"bytes"
"encoding/json"
"fmt"
"io"
"os"
"sync"
"sync/atomic"
"golang.org/x/net/websocket"
"github.com/zrepl/zrepl/internal/util/envconst"
)
var chrometracePID string
func init() {
var err error
chrometracePID, err = os.Hostname()
if err != nil {
panic(err)
}
chrometracePID = fmt.Sprintf("%q", chrometracePID)
}
type chrometraceEvent struct {
Cat string `json:"cat,omitempty"`
Name string `json:"name"`
Stack []string `json:"stack,omitempty"`
Phase string `json:"ph"`
TimestampUnixMicroseconds int64 `json:"ts"`
DurationMicroseconds int64 `json:"dur,omitempty"`
Pid string `json:"pid"`
Tid string `json:"tid"`
Id string `json:"id,omitempty"`
}
func chrometraceBeginSpan(s *traceNode) {
taskName := s.TaskName()
chrometraceWrite(chrometraceEvent{
Name: s.annotation,
Phase: "B",
TimestampUnixMicroseconds: s.startedAt.UnixNano() / 1000,
Pid: chrometracePID,
Tid: taskName,
})
}
func chrometraceEndSpan(s *traceNode) {
taskName := s.TaskName()
chrometraceWrite(chrometraceEvent{
Name: s.annotation,
Phase: "E",
TimestampUnixMicroseconds: s.endedAt.UnixNano() / 1000,
Pid: chrometracePID,
Tid: taskName,
})
}
var chrometraceFlowId uint64
func chrometraceBeginTask(s *traceNode) {
chrometraceBeginSpan(s)
if s.parentTask == nil {
return
}
// beginning of a task that has a parent
// => use flow events to link parent and child
flowId := atomic.AddUint64(&chrometraceFlowId, 1)
flowIdStr := fmt.Sprintf("%x", flowId)
parentTask := s.parentTask.TaskName()
chrometraceWrite(chrometraceEvent{
Cat: "task", // seems to be necessary, otherwise the GUI shows some `indexOf` JS error
Name: "child-task",
Phase: "s",
TimestampUnixMicroseconds: s.startedAt.UnixNano() / 1000, // yes, the child's timestamp (=> from-point of the flow line is at right x-position of parent's bar)
Pid: chrometracePID,
Tid: parentTask,
Id: flowIdStr,
})
childTask := s.TaskName()
if parentTask == childTask {
panic(parentTask)
}
chrometraceWrite(chrometraceEvent{
Cat: "task", // seems to be necessary, otherwise the GUI shows some `indexOf` JS error
Name: "child-task",
Phase: "f",
TimestampUnixMicroseconds: s.startedAt.UnixNano() / 1000,
Pid: chrometracePID,
Tid: childTask,
Id: flowIdStr,
})
}
func chrometraceEndTask(s *traceNode) {
chrometraceEndSpan(s)
}
type chrometraceConsumerRegistration struct {
w io.Writer
// errored must have capacity 1, the writer thread will send to it non-blocking, then close it
errored chan error
}
var chrometraceConsumers struct {
register chan chrometraceConsumerRegistration
consumers map[chrometraceConsumerRegistration]bool
write chan []byte
}
func init() {
chrometraceConsumers.register = make(chan chrometraceConsumerRegistration)
chrometraceConsumers.consumers = make(map[chrometraceConsumerRegistration]bool)
chrometraceConsumers.write = make(chan []byte)
go func() {
kickConsumer := func(c chrometraceConsumerRegistration, err error) {
debug("chrometrace kicking consumer %#v after error %v", c, err)
select {
case c.errored <- err:
default:
}
close(c.errored)
delete(chrometraceConsumers.consumers, c)
}
for {
select {
case reg := <-chrometraceConsumers.register:
debug("registered chrometrace consumer %#v", reg)
chrometraceConsumers.consumers[reg] = true
n, err := reg.w.Write([]byte("[\n"))
if err != nil {
kickConsumer(reg, err)
} else if n != 2 {
kickConsumer(reg, fmt.Errorf("short write: %v", n))
}
// successfully registered
case buf := <-chrometraceConsumers.write:
debug("chrometrace write request: %s", string(buf))
var r bytes.Reader
for c := range chrometraceConsumers.consumers {
r.Reset(buf)
n, err := io.Copy(c.w, &r)
debug("chrometrace wrote n=%v bytes to consumer %#v", n, c)
if err != nil {
kickConsumer(c, err)
}
}
}
}
}()
}
func chrometraceWrite(i interface{}) {
var buf bytes.Buffer
err := json.NewEncoder(&buf).Encode(i)
if err != nil {
panic(err)
}
buf.WriteString(",")
chrometraceConsumers.write <- buf.Bytes()
}
func ChrometraceClientWebsocketHandler(conn *websocket.Conn) {
defer conn.Close()
var wg sync.WaitGroup
defer wg.Wait()
wg.Add(1)
go func() {
defer wg.Done()
r := bufio.NewReader(conn)
_, _, _ = r.ReadLine() // ignore errors
conn.Close()
}()
errored := make(chan error, 1)
chrometraceConsumers.register <- chrometraceConsumerRegistration{
w: conn,
errored: errored,
}
wg.Add(1)
go func() {
defer wg.Done()
<-errored
conn.Close()
}()
}
var chrometraceFileConsumerPath = envconst.String("ZREPL_ACTIVITY_TRACE", "")
func init() {
if chrometraceFileConsumerPath != "" {
var err error
f, err := os.Create(chrometraceFileConsumerPath)
if err != nil {
panic(err)
}
errored := make(chan error, 1)
chrometraceConsumers.register <- chrometraceConsumerRegistration{
w: f,
errored: errored,
}
go func() {
<-errored
f.Close()
}()
}
}
@@ -0,0 +1,27 @@
package trace
import "context"
type contextKey int
const (
contextKeyTraceNode contextKey = 1 + iota
)
var contextKeys = []contextKey{
contextKeyTraceNode,
}
// WithInherit inherits the task hierarchy from inheritFrom into ctx.
// The returned context is a child of ctx, but its task and span are those of inheritFrom.
//
// Note that in most use cases, callers most likely want to call WithTask since it will most likely
// be in some sort of connection handler context.
func WithInherit(ctx, inheritFrom context.Context) context.Context {
for _, k := range contextKeys {
if v := inheritFrom.Value(k); v != nil {
ctx = context.WithValue(ctx, k, v) // no shadow
}
}
return ctx
}
@@ -0,0 +1,76 @@
package trace
import (
"context"
"fmt"
"runtime"
"strings"
"sync"
)
// use like this:
//
// defer WithSpanFromStackUpdateCtx(&existingCtx)()
func WithSpanFromStackUpdateCtx(ctx *context.Context) DoneFunc {
childSpanCtx, end := WithSpan(*ctx, getMyCallerOrPanic())
*ctx = childSpanCtx
return end
}
// derive task name from call stack (caller's name)
func WithTaskFromStack(ctx context.Context) (context.Context, DoneFunc) {
return WithTask(ctx, getMyCallerOrPanic())
}
// derive task name from call stack (caller's name) and update *ctx
// to point to be the child task ctx
func WithTaskFromStackUpdateCtx(ctx *context.Context) DoneFunc {
child, end := WithTask(*ctx, getMyCallerOrPanic())
*ctx = child
return end
}
// create a task and a span within it in one call
func WithTaskAndSpan(ctx context.Context, task string, span string) (context.Context, DoneFunc) {
ctx, endTask := WithTask(ctx, task)
ctx, endSpan := WithSpan(ctx, fmt.Sprintf("%s %s", task, span))
return ctx, func() {
endSpan()
endTask()
}
}
// create a span during which several child tasks are spawned using the `add` function
//
// IMPORTANT FOR USERS: Caller must ensure that the capturing behavior is correct, the Go linter doesn't catch this.
func WithTaskGroup(ctx context.Context, taskGroup string) (_ context.Context, add func(f func(context.Context)), waitEnd DoneFunc) {
var wg sync.WaitGroup
ctx, endSpan := WithSpan(ctx, taskGroup)
add = func(f func(context.Context)) {
wg.Add(1)
go func() {
defer wg.Done()
ctx, endTask := WithTask(ctx, taskGroup)
defer endTask()
f(ctx)
}()
}
waitEnd = func() {
wg.Wait()
endSpan()
}
return ctx, add, waitEnd
}
func getMyCallerOrPanic() string {
pc, _, _, ok := runtime.Caller(2)
if !ok {
panic("cannot get caller")
}
details := runtime.FuncForPC(pc)
if ok && details != nil {
const prefix = "github.com/zrepl/zrepl"
return strings.TrimPrefix(strings.TrimPrefix(details.Name(), prefix), "/")
}
return ""
}
@@ -0,0 +1,62 @@
package trace
import (
"context"
"sync/atomic"
"testing"
"time"
"github.com/stretchr/testify/assert"
)
func TestGetCallerOrPanic(t *testing.T) {
withStackFromCtxMock := func() string {
return getMyCallerOrPanic()
}
ret := withStackFromCtxMock()
// zrepl prefix is stripped
assert.Equal(t, "internal/daemon/logging/trace.TestGetCallerOrPanic", ret)
}
func TestWithTaskGroupRunTasksConcurrently(t *testing.T) {
// spawn a task group where each task waits for the other to start
// => without concurrency, they would hang
rootCtx, endRoot := WithTaskFromStack(context.Background())
defer endRoot()
_, add, waitEnd := WithTaskGroup(rootCtx, "test-task-group")
schedulerTimeout := 2 * time.Second
timeout := time.After(schedulerTimeout)
var hadTimeout uint32
started0, started1 := make(chan struct{}), make(chan struct{})
for i := 0; i < 2; i++ {
i := i // capture by copy
add(func(ctx context.Context) {
switch i {
case 0:
close(started0)
select {
case <-started1:
case <-timeout:
atomic.AddUint32(&hadTimeout, 1)
}
case 1:
close(started1)
select {
case <-started0:
case <-timeout:
atomic.AddUint32(&hadTimeout, 1)
}
default:
panic("unreachable")
}
})
}
waitEnd()
assert.Zero(t, hadTimeout, "either bad impl or scheduler timeout (which is %v)", schedulerTimeout)
}
@@ -0,0 +1,19 @@
package trace
import (
"fmt"
"os"
"github.com/zrepl/zrepl/internal/util/envconst"
)
const debugEnabledEnvVar = "ZREPL_TRACE_DEBUG_ENABLED"
var debugEnabled = envconst.Bool(debugEnabledEnvVar, false)
func debug(format string, args ...interface{}) {
if !debugEnabled {
return
}
fmt.Fprintf(os.Stderr, format+"\n", args...)
}
@@ -0,0 +1,35 @@
package trace
import (
"crypto/rand"
"encoding/base64"
"strings"
"github.com/zrepl/zrepl/internal/util/envconst"
)
var genIdNumBytes = envconst.Int("ZREPL_TRACE_ID_NUM_BYTES", 3)
func init() {
if genIdNumBytes < 1 {
panic("trace node id byte length must be at least 1")
}
}
func genID() string {
var out strings.Builder
enc := base64.NewEncoder(base64.RawStdEncoding, &out)
buf := make([]byte, genIdNumBytes)
_, err := rand.Read(buf)
if err != nil {
panic(err)
}
n, err := enc.Write(buf[:])
if err != nil || n != len(buf) {
panic(err)
}
if err := enc.Close(); err != nil {
panic(err)
}
return out.String()
}
+205
View File
@@ -0,0 +1,205 @@
package trace
import (
"context"
"fmt"
"testing"
"github.com/gitchander/permutation"
"github.com/pkg/errors"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
func TestRegularSpanUsage(t *testing.T) {
root, endRoot := WithTask(context.Background(), "root")
defer endRoot()
s1, endS1 := WithSpan(root, "parent")
s2, endS2 := WithSpan(s1, "child")
_, endS3 := WithSpan(s2, "grand-child")
require.NotPanics(t, func() { endS3() })
require.NotPanics(t, func() { endS2() })
// reuse
_, endS4 := WithSpan(s1, "child-2")
require.NotPanics(t, func() { endS4() })
// close parent
require.NotPanics(t, func() { endS1() })
}
func TestMultipleActiveChildSpansNotAllowed(t *testing.T) {
root, endRoot := WithTask(context.Background(), "root")
defer endRoot()
s1, _ := WithSpan(root, "s1")
_, endS2 := WithSpan(s1, "s1-child1")
require.PanicsWithValue(t, ErrAlreadyActiveChildSpan, func() {
_, _ = WithSpan(s1, "s1-child2")
})
endS2()
require.NotPanics(t, func() {
_, _ = WithSpan(s1, "s1-child2")
})
}
func TestForkingChildSpansNotAllowed(t *testing.T) {
root, endRoot := WithTask(context.Background(), "root")
defer endRoot()
s1, _ := WithSpan(root, "s1")
sc, endSC := WithSpan(s1, "s1-child")
_, _ = WithSpan(sc, "s1-child-child")
require.PanicsWithValue(t, ErrSpanStillHasActiveChildSpan, func() {
endSC()
})
}
func TestRegularTaskUsage(t *testing.T) {
// assert concurrent activities on different tasks can end in any order
closeOrder := []int{0, 1, 2}
closeOrders := permutation.New(permutation.IntSlice(closeOrder))
for closeOrders.Next() {
t.Run(fmt.Sprintf("%v", closeOrder), func(t *testing.T) {
root, endRoot := WithTask(context.Background(), "root")
defer endRoot()
c1, endC1 := WithTask(root, "c1")
defer endC1()
c2, endC2 := WithTask(root, "c2")
defer endC2()
// begin 3 concurrent activities
_, endAR := WithSpan(root, "aR")
_, endAC1 := WithSpan(c1, "aC1")
_, endAC2 := WithSpan(c2, "aC2")
endFuncs := []DoneFunc{endAR, endAC1, endAC2}
for _, i := range closeOrder {
require.NotPanics(t, func() {
endFuncs[i]()
}, "%v", i)
}
})
}
}
func TestTaskEndWithActiveChildTaskNotAllowed(t *testing.T) {
root, _ := WithTask(context.Background(), "root")
c, endC := WithTask(root, "child")
_, _ = WithTask(c, "grand-child")
func() {
defer func() {
r := recover()
require.NotNil(t, r)
err, ok := r.(error)
require.True(t, ok)
require.Equal(t, ErrTaskStillHasActiveChildTasks, errors.Cause(err))
}()
endC()
}()
}
func TestIdempotentEndTask(t *testing.T) {
_, end := WithTask(context.Background(), "root")
end()
require.NotPanics(t, func() { end() })
}
func TestSpansPanicIfNoParentTask(t *testing.T) {
require.Panics(t, func() { WithSpan(context.Background(), "taskless-span") })
}
func TestIdempotentEndSpan(t *testing.T) {
root, _ := WithTask(context.Background(), "root")
_, end := WithSpan(root, "span")
end()
require.NotPanics(t, func() { end() })
}
func logAndGetTraceNode(t *testing.T, descr string, ctx context.Context) *traceNode {
n, ok := ctx.Value(contextKeyTraceNode).(*traceNode)
require.True(t, ok)
t.Logf("% 20s %p %#v", descr, n, n)
return n
}
func TestWhiteboxHierachy(t *testing.T) {
root, e1 := WithTask(context.Background(), "root")
rootN := logAndGetTraceNode(t, "root", root)
assert.Nil(t, rootN.parentTask)
assert.Nil(t, rootN.parentSpan)
child, e2 := WithSpan(root, "child")
childN := logAndGetTraceNode(t, "child", child)
assert.Equal(t, rootN, childN.parentTask)
assert.Equal(t, rootN, childN.parentSpan)
grandchild, e3 := WithSpan(child, "grandchild")
grandchildN := logAndGetTraceNode(t, "grandchild", grandchild)
assert.Equal(t, rootN, grandchildN.parentTask)
assert.Equal(t, childN, grandchildN.parentSpan)
gcTask, e4 := WithTask(grandchild, "grandchild-task")
gcTaskN := logAndGetTraceNode(t, "grandchild-task", gcTask)
assert.Equal(t, rootN, gcTaskN.parentTask)
assert.Nil(t, gcTaskN.parentSpan)
// it is allowed that a child task outlives the _span_ in which it was created
// (albeit not its parent task)
e3()
e2()
gcTaskSpan, e5 := WithSpan(gcTask, "granschild-task-span")
gcTaskSpanN := logAndGetTraceNode(t, "granschild-task-span", gcTaskSpan)
assert.Equal(t, gcTaskN, gcTaskSpanN.parentTask)
assert.Equal(t, gcTaskN, gcTaskSpanN.parentSpan)
e5()
e4()
e1()
}
func TestOrphanTasksBecomeNewRootWhenAllAncestorsAreDead(t *testing.T) {
parent, e1 := WithTask(context.Background(), "parent")
_ = logAndGetTraceNode(t, "parent-task", parent)
e1()
var child context.Context
var e2 DoneFunc
require.NotPanics(t, func() {
child, e2 = WithTask(parent, "child")
})
childN := logAndGetTraceNode(t, "child-task", child)
assert.Nil(t, childN.parentTask)
grandchild, _ := WithTask(child, "grandchild")
grandchildN := logAndGetTraceNode(t, "grandchild-task", grandchild)
assert.Equal(t, childN, grandchildN.parentTask)
require.Panics(t, func() { e2() }, "if the parent was alive at creation of child task, wrong termination order remains a panicable offense though")
}
func TestOrphanTaskBecomesChildToNearestLiveAncestor(t *testing.T) {
parent, e1 := WithTask(context.Background(), "parent")
parentN := logAndGetTraceNode(t, "parent-task", parent)
child, e2 := WithTask(parent, "child")
childN := logAndGetTraceNode(t, "child-task", child)
assert.Equal(t, parentN, childN.parentTask)
e2()
grandchild, e3 := WithTask(child, "grandchild")
grandchildN := logAndGetTraceNode(t, "grandchild-task", grandchild)
assert.Equal(t, parentN, grandchildN.parentTask) // child is already dead
require.NotPanics(t, func() {
e3()
e1()
})
}
@@ -0,0 +1,53 @@
package trace
import (
"fmt"
"strings"
"sync"
"github.com/bits-and-blooms/bitset"
)
type uniqueConcurrentTaskNamer struct {
mtx sync.Mutex
active map[string]*bitset.BitSet
}
// bitvecLengthGauge may be nil
func newUniqueTaskNamer() *uniqueConcurrentTaskNamer {
return &uniqueConcurrentTaskNamer{
active: make(map[string]*bitset.BitSet),
}
}
// appends `#%d` to `name` such that until `done` is called,
// it is guaranteed that `#%d` is not returned a second time for the same `name`
func (namer *uniqueConcurrentTaskNamer) UniqueConcurrentTaskName(name string) (uniqueName string, done func()) {
if strings.Contains(name, "#") {
panic(name)
}
namer.mtx.Lock()
act, ok := namer.active[name]
if !ok {
act = bitset.New(64) // FIXME magic const
namer.active[name] = act
}
id, ok := act.NextClear(0)
if !ok {
// if !ok, all bits are 1 and act.Len() returns the next bit
id = act.Len()
// FIXME unbounded growth without reclamation
}
act.Set(id)
namer.mtx.Unlock()
return fmt.Sprintf("%s#%d", name, id), func() {
namer.mtx.Lock()
defer namer.mtx.Unlock()
act, ok := namer.active[name]
if !ok {
panic("must be initialized upon entry")
}
act.Clear(id)
}
}
@@ -0,0 +1,58 @@
package trace
import (
"fmt"
"sync"
"sync/atomic"
"testing"
"github.com/bits-and-blooms/bitset"
"github.com/stretchr/testify/require"
)
func TestBitsetFeaturesForUniqueConcurrentTaskNamer(t *testing.T) {
var b bitset.BitSet
require.Equal(t, uint(0), b.Len())
require.Equal(t, uint(0), b.Count())
b.Set(0)
require.Equal(t, uint(1), b.Len())
require.Equal(t, uint(1), b.Count())
b.Set(8)
require.Equal(t, uint(9), b.Len())
require.Equal(t, uint(2), b.Count())
b.Set(1)
require.Equal(t, uint(9), b.Len())
require.Equal(t, uint(3), b.Count())
}
func TestUniqueConcurrentTaskNamer(t *testing.T) {
namer := newUniqueTaskNamer()
var wg sync.WaitGroup
const N = 8128
const Q = 23
var fails uint32
var m sync.Map
wg.Add(N)
for i := 0; i < N; i++ {
go func(i int) {
defer wg.Done()
name := fmt.Sprintf("%d", i/Q)
uniqueName, done := namer.UniqueConcurrentTaskName(name)
act, _ := m.LoadOrStore(uniqueName, i)
if act.(int) != i {
atomic.AddUint32(&fails, 1)
}
m.Delete(uniqueName)
done()
}(i)
}
wg.Wait()
require.Equal(t, uint32(0), fails)
}