mirror of
https://github.com/VictoriaMetrics/VictoriaMetrics.git
synced 2024-12-30 07:40:06 +01:00
178 lines
5.6 KiB
Go
178 lines
5.6 KiB
Go
|
/*
|
||
|
Copyright 2014 The Kubernetes Authors.
|
||
|
|
||
|
Licensed under the Apache License, Version 2.0 (the "License");
|
||
|
you may not use this file except in compliance with the License.
|
||
|
You may obtain a copy of the License at
|
||
|
|
||
|
http://www.apache.org/licenses/LICENSE-2.0
|
||
|
|
||
|
Unless required by applicable law or agreed to in writing, software
|
||
|
distributed under the License is distributed on an "AS IS" BASIS,
|
||
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||
|
See the License for the specific language governing permissions and
|
||
|
limitations under the License.
|
||
|
*/
|
||
|
|
||
|
package runtime
|
||
|
|
||
|
import (
|
||
|
"fmt"
|
||
|
"net/http"
|
||
|
"runtime"
|
||
|
"sync"
|
||
|
"time"
|
||
|
|
||
|
"k8s.io/klog/v2"
|
||
|
)
|
||
|
|
||
|
var (
|
||
|
// ReallyCrash controls the behavior of HandleCrash and defaults to
|
||
|
// true. It's exposed so components can optionally set to false
|
||
|
// to restore prior behavior. This flag is mostly used for tests to validate
|
||
|
// crash conditions.
|
||
|
ReallyCrash = true
|
||
|
)
|
||
|
|
||
|
// PanicHandlers is a list of functions which will be invoked when a panic happens.
|
||
|
var PanicHandlers = []func(interface{}){logPanic}
|
||
|
|
||
|
// HandleCrash simply catches a crash and logs an error. Meant to be called via
|
||
|
// defer. Additional context-specific handlers can be provided, and will be
|
||
|
// called in case of panic. HandleCrash actually crashes, after calling the
|
||
|
// handlers and logging the panic message.
|
||
|
//
|
||
|
// E.g., you can provide one or more additional handlers for something like shutting down go routines gracefully.
|
||
|
func HandleCrash(additionalHandlers ...func(interface{})) {
|
||
|
if r := recover(); r != nil {
|
||
|
for _, fn := range PanicHandlers {
|
||
|
fn(r)
|
||
|
}
|
||
|
for _, fn := range additionalHandlers {
|
||
|
fn(r)
|
||
|
}
|
||
|
if ReallyCrash {
|
||
|
// Actually proceed to panic.
|
||
|
panic(r)
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// logPanic logs the caller tree when a panic occurs (except in the special case of http.ErrAbortHandler).
|
||
|
func logPanic(r interface{}) {
|
||
|
if r == http.ErrAbortHandler {
|
||
|
// honor the http.ErrAbortHandler sentinel panic value:
|
||
|
// ErrAbortHandler is a sentinel panic value to abort a handler.
|
||
|
// While any panic from ServeHTTP aborts the response to the client,
|
||
|
// panicking with ErrAbortHandler also suppresses logging of a stack trace to the server's error log.
|
||
|
return
|
||
|
}
|
||
|
|
||
|
// Same as stdlib http server code. Manually allocate stack trace buffer size
|
||
|
// to prevent excessively large logs
|
||
|
const size = 64 << 10
|
||
|
stacktrace := make([]byte, size)
|
||
|
stacktrace = stacktrace[:runtime.Stack(stacktrace, false)]
|
||
|
if _, ok := r.(string); ok {
|
||
|
klog.Errorf("Observed a panic: %s\n%s", r, stacktrace)
|
||
|
} else {
|
||
|
klog.Errorf("Observed a panic: %#v (%v)\n%s", r, r, stacktrace)
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// ErrorHandlers is a list of functions which will be invoked when a nonreturnable
|
||
|
// error occurs.
|
||
|
// TODO(lavalamp): for testability, this and the below HandleError function
|
||
|
// should be packaged up into a testable and reusable object.
|
||
|
var ErrorHandlers = []func(error){
|
||
|
logError,
|
||
|
(&rudimentaryErrorBackoff{
|
||
|
lastErrorTime: time.Now(),
|
||
|
// 1ms was the number folks were able to stomach as a global rate limit.
|
||
|
// If you need to log errors more than 1000 times a second you
|
||
|
// should probably consider fixing your code instead. :)
|
||
|
minPeriod: time.Millisecond,
|
||
|
}).OnError,
|
||
|
}
|
||
|
|
||
|
// HandlerError is a method to invoke when a non-user facing piece of code cannot
|
||
|
// return an error and needs to indicate it has been ignored. Invoking this method
|
||
|
// is preferable to logging the error - the default behavior is to log but the
|
||
|
// errors may be sent to a remote server for analysis.
|
||
|
func HandleError(err error) {
|
||
|
// this is sometimes called with a nil error. We probably shouldn't fail and should do nothing instead
|
||
|
if err == nil {
|
||
|
return
|
||
|
}
|
||
|
|
||
|
for _, fn := range ErrorHandlers {
|
||
|
fn(err)
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// logError prints an error with the call stack of the location it was reported
|
||
|
func logError(err error) {
|
||
|
klog.ErrorDepth(2, err)
|
||
|
}
|
||
|
|
||
|
type rudimentaryErrorBackoff struct {
|
||
|
minPeriod time.Duration // immutable
|
||
|
// TODO(lavalamp): use the clock for testability. Need to move that
|
||
|
// package for that to be accessible here.
|
||
|
lastErrorTimeLock sync.Mutex
|
||
|
lastErrorTime time.Time
|
||
|
}
|
||
|
|
||
|
// OnError will block if it is called more often than the embedded period time.
|
||
|
// This will prevent overly tight hot error loops.
|
||
|
func (r *rudimentaryErrorBackoff) OnError(error) {
|
||
|
now := time.Now() // start the timer before acquiring the lock
|
||
|
r.lastErrorTimeLock.Lock()
|
||
|
d := now.Sub(r.lastErrorTime)
|
||
|
r.lastErrorTime = time.Now()
|
||
|
r.lastErrorTimeLock.Unlock()
|
||
|
|
||
|
// Do not sleep with the lock held because that causes all callers of HandleError to block.
|
||
|
// We only want the current goroutine to block.
|
||
|
// A negative or zero duration causes time.Sleep to return immediately.
|
||
|
// If the time moves backwards for any reason, do nothing.
|
||
|
time.Sleep(r.minPeriod - d)
|
||
|
}
|
||
|
|
||
|
// GetCaller returns the caller of the function that calls it.
|
||
|
func GetCaller() string {
|
||
|
var pc [1]uintptr
|
||
|
runtime.Callers(3, pc[:])
|
||
|
f := runtime.FuncForPC(pc[0])
|
||
|
if f == nil {
|
||
|
return "Unable to find caller"
|
||
|
}
|
||
|
return f.Name()
|
||
|
}
|
||
|
|
||
|
// RecoverFromPanic replaces the specified error with an error containing the
|
||
|
// original error, and the call tree when a panic occurs. This enables error
|
||
|
// handlers to handle errors and panics the same way.
|
||
|
func RecoverFromPanic(err *error) {
|
||
|
if r := recover(); r != nil {
|
||
|
// Same as stdlib http server code. Manually allocate stack trace buffer size
|
||
|
// to prevent excessively large logs
|
||
|
const size = 64 << 10
|
||
|
stacktrace := make([]byte, size)
|
||
|
stacktrace = stacktrace[:runtime.Stack(stacktrace, false)]
|
||
|
|
||
|
*err = fmt.Errorf(
|
||
|
"recovered from panic %q. (err=%v) Call stack:\n%s",
|
||
|
r,
|
||
|
*err,
|
||
|
stacktrace)
|
||
|
}
|
||
|
}
|
||
|
|
||
|
// Must panics on non-nil errors. Useful to handling programmer level errors.
|
||
|
func Must(err error) {
|
||
|
if err != nil {
|
||
|
panic(err)
|
||
|
}
|
||
|
}
|