volcano.sh/volcano@v1.9.0/pkg/scheduler/cache/dumper.go (about)

     1  /*
     2   Copyright 2023 The Volcano Authors.
     3  
     4   Licensed under the Apache License, Version 2.0 (the "License");
     5   you may not use this file except in compliance with the License.
     6   You may obtain a copy of the License at
     7  
     8       http://www.apache.org/licenses/LICENSE-2.0
     9  
    10   Unless required by applicable law or agreed to in writing, software
    11   distributed under the License is distributed on an "AS IS" BASIS,
    12   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    13   See the License for the specific language governing permissions and
    14   limitations under the License.
    15  */
    16  
    17  package cache
    18  
    19  import (
    20  	"encoding/json"
    21  	"fmt"
    22  	"os"
    23  	"os/signal"
    24  	"path"
    25  	"runtime"
    26  	"strings"
    27  	"syscall"
    28  	"time"
    29  
    30  	"k8s.io/klog/v2"
    31  
    32  	"volcano.sh/volcano/pkg/scheduler/api"
    33  )
    34  
    35  // Dumper writes some information from the scheduler cache to the scheduler logs
    36  // for debugging purposes. Usage: run `kill -s USR2 <pid>` in the shell, where <pid>
    37  // is the process id of the scheduler process.
    38  type Dumper struct {
    39  	Cache   Cache
    40  	RootDir string // target directory for the dumped json file
    41  }
    42  
    43  // dumpToJSONFile marsh scheduler cache snapshot to json file
    44  func (d *Dumper) dumpToJSONFile() {
    45  	snapshot := d.Cache.Snapshot()
    46  	name := fmt.Sprintf("snapshot-%d.json", time.Now().Unix())
    47  	fName := path.Join(d.RootDir, name)
    48  	file, err := os.OpenFile(fName, os.O_RDWR|os.O_CREATE|os.O_TRUNC, 0644)
    49  	if err != nil {
    50  		klog.Errorf("error creating snapshot because of error creating file: %v", err)
    51  		return
    52  	}
    53  	defer file.Close()
    54  	klog.Infoln("Starting to dump info in scheduler cache to file", fName)
    55  	if err = json.NewEncoder(file).Encode(snapshot.Nodes); err != nil {
    56  		klog.Errorf("Failed to dump info in scheduler cache, json encode error: %v", err)
    57  		return
    58  	}
    59  
    60  	klog.Infoln("Successfully dump info in scheduler cache to file", fName)
    61  }
    62  
    63  // dumpAll prints all information to log
    64  func (d *Dumper) dumpAll() {
    65  	snapshot := d.Cache.Snapshot()
    66  	klog.Info("Dump of nodes info in scheduler cache")
    67  	for _, nodeInfo := range snapshot.Nodes {
    68  		klog.Info(d.printNodeInfo(nodeInfo))
    69  	}
    70  
    71  	klog.Info("Dump of jobs info in scheduler cache")
    72  	for _, jobInfo := range snapshot.Jobs {
    73  		klog.Info(d.printJobInfo(jobInfo))
    74  	}
    75  
    76  	d.displaySchedulerMemStats()
    77  }
    78  
    79  func (d *Dumper) displaySchedulerMemStats() {
    80  	var m runtime.MemStats
    81  	runtime.ReadMemStats(&m)
    82  	klog.Infof("volcano scheduler memory stat: %+v\n", m)
    83  }
    84  
    85  func (d *Dumper) printNodeInfo(node *api.NodeInfo) string {
    86  	var data strings.Builder
    87  	data.WriteString("\n")
    88  	data.WriteString(node.String())
    89  	data.WriteString("\n")
    90  	return data.String()
    91  }
    92  
    93  func (d *Dumper) printJobInfo(jobInfo *api.JobInfo) string {
    94  	var data strings.Builder
    95  	data.WriteString("\n")
    96  	data.WriteString(jobInfo.String())
    97  	data.WriteString("\n")
    98  	return data.String()
    99  }
   100  
   101  // ListenForSignal starts a goroutine that will respond when process
   102  // receives SIGUSER1/SIGUSER2 signal.
   103  func (d *Dumper) ListenForSignal(stopCh <-chan struct{}) {
   104  	ch1 := make(chan os.Signal, 1)
   105  	ch2 := make(chan os.Signal, 1)
   106  	signal.Notify(ch1, syscall.SIGUSR1)
   107  	signal.Notify(ch2, syscall.SIGUSR2)
   108  	go func() {
   109  		for {
   110  			select {
   111  			case <-stopCh:
   112  				return
   113  			case <-ch1:
   114  				d.dumpToJSONFile()
   115  			case <-ch2:
   116  				d.dumpAll()
   117  			}
   118  		}
   119  	}()
   120  }