mirror of
https://codeberg.org/forgejo/forgejo.git
synced 2024-11-15 06:26:16 +01:00
c88547ce71
Continues on from #19202. Following the addition of pprof labels we can now more easily understand the relationship between a goroutine and the requests that spawn them. This PR takes advantage of the labels and adds a few others, then provides a mechanism for the monitoring page to query the pprof goroutine profile. The binary profile that results from this profile is immediately piped in to the google library for parsing this and then stack traces are formed for the goroutines. If the goroutine is within a context or has been created from a goroutine within a process context it will acquire the process description labels for that process. The goroutines are mapped with there associate pids and any that do not have an associated pid are placed in a group at the bottom as unbound. In this way we should be able to more easily examine goroutines that have been stuck. A manager command `gitea manager processes` is also provided that can export the processes (with or without stacktraces) to the command line. Signed-off-by: Andrew Thornton <art27@cantab.net>
125 lines
3.4 KiB
Go
125 lines
3.4 KiB
Go
// Copyright 2021 The Gitea Authors. All rights reserved.
|
|
// Use of this source code is governed by a MIT-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package routing
|
|
|
|
import (
|
|
"context"
|
|
"net/http"
|
|
"sync"
|
|
"time"
|
|
|
|
"code.gitea.io/gitea/modules/graceful"
|
|
"code.gitea.io/gitea/modules/process"
|
|
)
|
|
|
|
// Event indicates when the printer is triggered
|
|
type Event int
|
|
|
|
const (
|
|
// StartEvent at the beginning of a request
|
|
StartEvent Event = iota
|
|
|
|
// StillExecutingEvent the request is still executing
|
|
StillExecutingEvent
|
|
|
|
// EndEvent the request has ended (either completed or failed)
|
|
EndEvent
|
|
)
|
|
|
|
// Printer is used to output the log for a request
|
|
type Printer func(trigger Event, record *requestRecord)
|
|
|
|
type requestRecordsManager struct {
|
|
print Printer
|
|
|
|
lock sync.Mutex
|
|
|
|
requestRecords map[uint64]*requestRecord
|
|
count uint64
|
|
}
|
|
|
|
func (manager *requestRecordsManager) startSlowQueryDetector(threshold time.Duration) {
|
|
go graceful.GetManager().RunWithShutdownContext(func(ctx context.Context) {
|
|
ctx, _, finished := process.GetManager().AddTypedContext(ctx, "Service: SlowQueryDetector", process.SystemProcessType, true)
|
|
defer finished()
|
|
// This go-routine checks all active requests every second.
|
|
// If a request has been running for a long time (eg: /user/events), we also print a log with "still-executing" message
|
|
// After the "still-executing" log is printed, the record will be removed from the map to prevent from duplicated logs in future
|
|
|
|
// We do not care about accurate duration here. It just does the check periodically, 0.5s or 1.5s are all OK.
|
|
t := time.NewTicker(time.Second)
|
|
for {
|
|
select {
|
|
case <-ctx.Done():
|
|
return
|
|
case <-t.C:
|
|
now := time.Now()
|
|
|
|
var slowRequests []*requestRecord
|
|
|
|
// find all slow requests with lock
|
|
manager.lock.Lock()
|
|
for index, record := range manager.requestRecords {
|
|
if now.Sub(record.startTime) < threshold {
|
|
continue
|
|
}
|
|
|
|
slowRequests = append(slowRequests, record)
|
|
delete(manager.requestRecords, index)
|
|
}
|
|
manager.lock.Unlock()
|
|
|
|
// print logs for slow requests
|
|
for _, record := range slowRequests {
|
|
manager.print(StillExecutingEvent, record)
|
|
}
|
|
}
|
|
}
|
|
})
|
|
}
|
|
|
|
func (manager *requestRecordsManager) handler(next http.Handler) http.Handler {
|
|
return http.HandlerFunc(func(w http.ResponseWriter, req *http.Request) {
|
|
record := &requestRecord{
|
|
startTime: time.Now(),
|
|
request: req,
|
|
responseWriter: w,
|
|
}
|
|
|
|
// generate a record index an insert into the map
|
|
manager.lock.Lock()
|
|
record.index = manager.count
|
|
manager.count++
|
|
manager.requestRecords[record.index] = record
|
|
manager.lock.Unlock()
|
|
|
|
defer func() {
|
|
// just in case there is a panic. now the panics are all recovered in middleware.go
|
|
localPanicErr := recover()
|
|
if localPanicErr != nil {
|
|
record.lock.Lock()
|
|
record.panicError = localPanicErr
|
|
record.lock.Unlock()
|
|
}
|
|
|
|
// remove from the record map
|
|
manager.lock.Lock()
|
|
delete(manager.requestRecords, record.index)
|
|
manager.lock.Unlock()
|
|
|
|
// log the end of the request
|
|
manager.print(EndEvent, record)
|
|
|
|
if localPanicErr != nil {
|
|
// the panic wasn't recovered before us, so we should pass it up, and let the framework handle the panic error
|
|
panic(localPanicErr)
|
|
}
|
|
}()
|
|
|
|
req = req.WithContext(context.WithValue(req.Context(), contextKey, record))
|
|
manager.print(StartEvent, record)
|
|
next.ServeHTTP(w, req)
|
|
})
|
|
}
|