failpoint

An implementation of failpoints for Golang. Fail points are used to add code points where errors may be injected in a user controlled fashion. Fail point is a code snippet that is only executed when the corresponding failpoint is active.

Quick Start (use failpoint-ctl)

1. Build failpoint-ctl from source

   git clone https://github.com/pingcap/failpoint.git
   cd failpoint
   make
   ls bin/failpoint-ctl
   

2. Inject failpoints to your program, eg:

   package main
   
   import "github.com/pingcap/failpoint"
   
   func main() {
       failpoint.Inject("testPanic", func() {
           panic("failpoint triggerd")
       })
   }
   

3. Transfrom your code with failpoint-ctl enable

4. Build with go build

5. Enable failpoints with GO_FAILPOINTS environment variable

   GO_FAILPOINTS="main/testPanic=return(true)" ./your-program
   

   Note: GO_FAILPOINTS does not work with InjectCall type of marker.

6. If you use go run to run the test, don't forget to add the generated binding__failpoint_binding__.go in your command, like:

   GO_FAILPOINTS="main/testPanic=return(true)" go run your-program.go binding__failpoint_binding__.go
   

Quick Start (use failpoint-toolexec)

1. Build failpoint-toolexec from source

   git clone https://github.com/pingcap/failpoint.git
   cd failpoint
   make
   ls bin/failpoint-toolexec
   

2. Inject failpoints to your program, eg:

   package main
   
   import "github.com/pingcap/failpoint"
   
   func main() {
       failpoint.Inject("testPanic", func() {
           panic("failpoint triggerd")
       })
   }
   

3. Use a separate build cache to avoid mixing caches without failpoint-toolexec, and build

   GOCACHE=/tmp/failpoint-cache go build -toolexec path/to/failpoint-toolexec

4. Enable failpoints with GO_FAILPOINTS environment variable

   GO_FAILPOINTS="main/testPanic=return(true)" ./your-program
   

5. You can also use go run or go test, like:

   GOCACHE=/tmp/failpoint-cache GO_FAILPOINTS="main/testPanic=return(true)" go run -toolexec path/to/failpoint-toolexec your-program.go
   

Design principles

• Define failpoint in valid Golang code, not comments or anything else

• Failpoint does not have any extra cost

  • Will not take effect on regular logic
  • Will not cause regular code performance regression
  • Failpoint code will not appear in the final binary

• Failpoint routine is writable/readable and should be checked by a compiler

• Generated code by failpoint definition is easy to read

• Keep the line numbers same with the injecting codes(easier to debug)

• Support parallel tests with context.Context

Key concepts

• Failpoint

  Faillpoint is a code snippet that is only executed when the corresponding failpoint is active. The closure will never be executed if failpoint.Disable("failpoint-name-for-demo") is executed.

  var outerVar = "declare in outer scope"
  failpoint.Inject("failpoint-name-for-demo", func(val failpoint.Value) {
      fmt.Println("unit-test", val, outerVar)
  })
  

• Marker functions

  • It is just an empty function

    • To hint the rewriter to rewrite with an equality statement
    • To receive some parameters as the rewrite rule
    • It will be inline in the compiling time and emit nothing to binary (zero cost)
    • The variables in external scope can be accessed in closure by capturing, and the converted code is still legal because all the captured-variables location in outer scope of IF statement.

  • It is easy to write/read

  • Introduce a compiler check for failpoints which cannot compile in the regular mode if failpoint code is invalid

• Marker funtion list

  • func Inject(fpname string, fpblock func(val Value)) {}
  • func InjectContext(fpname string, ctx context.Context, fpblock func(val Value)) {}
  • func InjectCall(fpname string, args ...any) {}
  • func Break(label ...string) {}
  • func Goto(label string) {}
  • func Continue(label ...string) {}
  • func Fallthrough() {}
  • func Return(results ...interface{}) {}
  • func Label(label string) {}

• Supported failpoint environment variable

  failpoint can be enabled by export environment variables with the following patten, which is quite similar to freebsd failpoint SYSCTL VARIABLES

  Note: InjectCall cannot be enabled by environment variables.

  [<percent>%][<count>*]<type>[(args...)][-><more terms>]
  

  The argument specifies which action to take; it can be one of:

  • off: Take no action (does not trigger failpoint code)
  • return: Trigger failpoint with specified argument
  • sleep: Sleep the specified number of milliseconds
  • panic: Panic
  • break: Execute gdb and break into debugger
  • print: Print failpoint path for inject variable
  • pause: Pause will pause until the failpoint is disabled

How to inject a failpoint to your program

• You can call failpoint.Inject to inject a failpoint to the call site, where failpoint-name is used to trigger the failpoint and failpoint-closure will be expanded as the body of the IF statement.

  failpoint.Inject("failpoint-name", func(val failpoint.Value) {
      failpoint.Return("unit-test", val)
  })
  

  The converted code looks like:

  if val, _err_ := failpoint.Eval(_curpkg_("failpoint-name")); _err_ == nil {
      return "unit-test", val
  }
  

• failpoint.Value is the value that passes by failpoint.Enable("failpoint-name", "return(5)") which can be ignored.

  failpoint.Inject("failpoint-name", func(_ failpoint.Value) {
      fmt.Println("unit-test")
  })
  

  OR

  failpoint.Inject("failpoint-name", func() {
      fmt.Println("unit-test")
  })
  

  And the converted code looks like:

  if _, _err_ := failpoint.Eval(_curpkg_("failpoint-name")); _err_ == nil {
      fmt.Println("unit-test")
  }
  

• Also, the failpoint closure can be a function which takes context.Context. You can do some customized things with context.Context like controlling whether a failpoint is active in parallel tests or other cases. For example,

  failpoint.InjectContext(ctx, "failpoint-name", func(val failpoint.Value) {
      fmt.Println("unit-test", val)
  })
  

  The converted code looks like:

  if val, _err_ := failpoint.EvalContext(ctx, _curpkg_("failpoint-name")); _err_ == nil {
      fmt.Println("unit-test", val)
  }
  

• You can ignore context.Context, and this will generate the same code as above non-context version. For example,

  failpoint.InjectContext(nil, "failpoint-name", func(val failpoint.Value) {
      fmt.Println("unit-test", val)
  })
  

  Becomes

  if val, _err_ := failpoint.EvalContext(nil, _curpkg_("failpoint-name")); _err_ == nil {
      fmt.Println("unit-test", val)
  }
  

• You can use failpoint.InjectCall to inject a function call, this type of marker can only be enabled using failpoint.EnableCall and it must be called in the same process as the InjectCall call site. Using this marker, you can avoid failpoint code pollute you source code. See examples.

• You can control a failpoint by failpoint.WithHook

  func (s *dmlSuite) TestCRUDParallel() {
      sctx := failpoint.WithHook(context.Backgroud(), func(ctx context.Context, fpname string) bool {
          return ctx.Value(fpname) != nil // Determine by ctx key
      })
      insertFailpoints = map[string]struct{} {
          "insert-record-fp": {},
          "insert-index-fp": {},
          "on-duplicate-fp": {},
      }
      ictx := failpoint.WithHook(context.Backgroud(), func(ctx context.Context, fpname string) bool {
          _, found := insertFailpoints[fpname] // Only enables some failpoints.
          return found
      })
      deleteFailpoints = map[string]struct{} {
          "tikv-is-busy-fp": {},
          "fetch-tso-timeout": {},
      }
      dctx := failpoint.WithHook(context.Backgroud(), func(ctx context.Context, fpname string) bool {
          _, found := deleteFailpoints[fpname] // Only disables failpoints. 
          return !found
      })
      // other DML parallel test cases.
      s.RunParallel(buildSelectTests(sctx))
      s.RunParallel(buildInsertTests(ictx))
      s.RunParallel(buildDeleteTests(dctx))
  }
  

• If you use a failpoint in the loop context, maybe you will use other marker functions.

  failpoint.Label("outer")
  for i := 0; i < 100; i++ {
      inner:
          for j := 0; j < 1000; j++ {
              switch rand.Intn(j) + i {
              case j / 5:
                  failpoint.Break()
              case j / 7:
                  failpoint.Continue("outer")
              case j / 9:
                  failpoint.Fallthrough()
              case j / 10:
                  failpoint.Goto("outer")
              default:
                  failpoint.Inject("failpoint-name", func(val failpoint.Value) {
                      fmt.Println("unit-test", val.(int))
                      if val == j/11 {
                          failpoint.Break("inner")
                      } else {
                          failpoint.Goto("outer")
                      }
                  })
          }
      }
  }
  

  The above code block will generate the following code:

  outer:
      for i := 0; i < 100; i++ {
      inner:
          for j := 0; j < 1000; j++ {
              switch rand.Intn(j) + i {
              case j / 5:
                  break
              case j / 7:
                  continue outer
              case j / 9:
                  fallthrough
              case j / 10:
                  goto outer
              default:
                  if val, _err_ := failpoint.Eval(_curpkg_("failpoint-name")); _err_ == nil {
                      fmt.Println("unit-test", val.(int))
                      if val == j/11 {
                          break inner
                      } else {
                          goto outer
                      }
                  }
              }
          }
      }
  

• You may doubt why we do not use label, break, continue, and fallthrough directly instead of using failpoint marker functions.

  • Any unused symbol like an ident or a label is not permitted in Golang. It will be invalid if some label is only used in the failpoint closure. For example,

    label1: // compiler error: unused label1
        failpoint.Inject("failpoint-name", func(val failpoint.Value) {
            if val.(int) == 1000 {
                goto label1 // illegal to use goto here
            }
            fmt.Println("unit-test", val)
        })
    

  • break and continue can only be used in the loop context, which is not legal in the Golang code if we use them in closure directly.

Some complicated failpoints demo

• Inject a failpoint to the IF INITIAL statement or CONDITIONAL expression

  if a, b := func() {
      failpoint.Inject("failpoint-name", func(val failpoint.Value) {
          fmt.Println("unit-test", val)
      })
  }, func() int { return rand.Intn(200) }(); b > func() int {
      failpoint.Inject("failpoint-name", func(val failpoint.Value) int {
          return val.(int)
      })
      return rand.Intn(3000)
  }() && b < func() int {
      failpoint.Inject("failpoint-name-2", func(val failpoint.Value) {
          return rand.Intn(val.(int))
      })
      return rand.Intn(6000)
  }() {
      a()
      failpoint.Inject("failpoint-name-3", func(val failpoint.Value) {
          fmt.Println("unit-test", val)
      })
  }
  

  The above code block will generate something like this:

  if a, b := func() {
      if val, _err_ := failpoint.Eval(_curpkg_("failpoint-name")); _err_ == nil {
          fmt.Println("unit-test", val)
      }
  }, func() int { return rand.Intn(200) }(); b > func() int {
      if val, _err_ := failpoint.Eval(_curpkg_("failpoint-name")); _err_ == nil {
          return val.(int)
      }
      return rand.Intn(3000)
  }() && b < func() int {
      if val, ok := failpoint.Eval(_curpkg_("failpoint-name-2")); ok {
          return rand.Intn(val.(int))
      }
      return rand.Intn(6000)
  }() {
      a()
      if val, ok := failpoint.Eval(_curpkg_("failpoint-name-3")); ok {
          fmt.Println("unit-test", val)
      }
  }
  

• Inject a failpoint to the SELECT statement to make it block one CASE if the failpoint is active

  func (s *StoreService) ExecuteStoreTask() {
      select {
      case <-func() chan *StoreTask {
          failpoint.Inject("priority-fp", func(_ failpoint.Value) {
              return make(chan *StoreTask)
          })
          return s.priorityHighCh
      }():
          fmt.Println("execute high priority task")
  
      case <- s.priorityNormalCh:
          fmt.Println("execute normal priority task")
  
      case <- s.priorityLowCh:
          fmt.Println("execute normal low task")
      }
  }
  

  The above code block will generate something like this:

  func (s *StoreService) ExecuteStoreTask() {
      select {
      case <-func() chan *StoreTask {
          if _, ok := failpoint.Eval(_curpkg_("priority-fp")); ok {
              return make(chan *StoreTask)
          })
          return s.priorityHighCh
      }():
          fmt.Println("execute high priority task")
  
      case <- s.priorityNormalCh:
          fmt.Println("execute normal priority task")
  
      case <- s.priorityLowCh:
          fmt.Println("execute normal low task")
      }
  }
  

• Inject a failpoint to dynamically extend SWITCH CASE arms

  switch opType := operator.Type(); {
  case opType == "balance-leader":
      fmt.Println("create balance leader steps")
  
  case opType == "balance-region":
      fmt.Println("create balance region steps")
  
  case opType == "scatter-region":
      fmt.Println("create scatter region steps")
  
  case func() bool {
      failpoint.Inject("dynamic-op-type", func(val failpoint.Value) bool {
          return strings.Contains(val.(string), opType)
      })
      return false
  }():
      fmt.Println("do something")
  
  default:
      panic("unsupported operator type")
  }
  

  The above code block will generate something like this:

  switch opType := operator.Type(); {
  case opType == "balance-leader":
      fmt.Println("create balance leader steps")
  
  case opType == "balance-region":
      fmt.Println("create balance region steps")
  
  case opType == "scatter-region":
      fmt.Println("create scatter region steps")
  
  case func() bool {
      if val, ok := failpoint.Eval(_curpkg_("dynamic-op-type")); ok {
          return strings.Contains(val.(string), opType)
      }
      return false
  }():
      fmt.Println("do something")
  
  default:
      panic("unsupported operator type")
  }
  

• More complicated failpoints

  • There are more complicated failpoint sites that can be injected to
    • for the loop INITIAL statement, CONDITIONAL expression and POST statement
    • for the RANGE statement
    • SWITCH INITIAL statement
    • …
  • Anywhere you can call a function

Failpoint name best practice

As you see above, _curpkg_ will automatically wrap the original failpoint name in failpoint.Eval call. You can think of _curpkg_ as a macro that automatically prepends the current package path to the failpoint name. For example,

package ddl // which parent package is `github.com/pingcap/tidb`

func demo() {
	// _curpkg_("the-original-failpoint-name") will be expanded as `github.com/pingcap/tidb/ddl/the-original-failpoint-name`
	if val, ok := failpoint.Eval(_curpkg_("the-original-failpoint-name")); ok {...}
}

You do not need to care about _curpkg_ in your application. It is automatically generated after running failpoint-ctl enable and is deleted with failpoint-ctl disable.

Because all failpoints in a package share the same namespace, we need to be careful to avoid name conflict. There are some recommended naming rules to improve this situation.

• Keep name unique in current subpackage

• Use a self-explanatory name for the failpoint

  You can enable failpoints by environment variables

  GO_FAILPOINTS="github.com/pingcap/tidb/ddl/renameTableErr=return(100);github.com/pingcap/tidb/planner/core/illegalPushDown=return(true);github.com/pingcap/pd/server/schedulers/balanceLeaderFailed=return(true)"
  

Implementation details

1. Define a group of marker functions
2. Parse imports and prune a source file which does not import a failpoint
3. Traverse AST to find marker function calls
4. Marker function calls will be rewritten with an IF statement, which calls failpoint.Eval to determine whether a failpoint is active and executes failpoint code if the failpoint is enabled

Acknowledgments

• Thanks gofail to provide initial implementation.