volcano.sh/volcano@v1.9.0/pkg/controllers/job/job_controller_resync.go (about)

     1  /*
     2  Copyright 2019 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 job
    18  
    19  import (
    20  	"context"
    21  	"fmt"
    22  	"time"
    23  
    24  	"golang.org/x/time/rate"
    25  	v1 "k8s.io/api/core/v1"
    26  	"k8s.io/apimachinery/pkg/api/errors"
    27  	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
    28  	"k8s.io/client-go/util/workqueue"
    29  	"k8s.io/klog/v2"
    30  )
    31  
    32  func newRateLimitingQueue() workqueue.RateLimitingInterface {
    33  	return workqueue.NewRateLimitingQueue(workqueue.NewMaxOfRateLimiter(
    34  		workqueue.NewItemExponentialFailureRateLimiter(5*time.Millisecond, 180*time.Second),
    35  		// 10 qps, 100 bucket size.  This is only for retry speed and its only the overall factor (not per item)
    36  		&workqueue.BucketRateLimiter{Limiter: rate.NewLimiter(rate.Limit(10), 100)},
    37  	))
    38  }
    39  
    40  func (cc *jobcontroller) processResyncTask() {
    41  	obj, shutdown := cc.errTasks.Get()
    42  	if shutdown {
    43  		return
    44  	}
    45  
    46  	// one task only resync 10 times
    47  	if cc.errTasks.NumRequeues(obj) > 10 {
    48  		cc.errTasks.Forget(obj)
    49  		return
    50  	}
    51  
    52  	defer cc.errTasks.Done(obj)
    53  
    54  	task, ok := obj.(*v1.Pod)
    55  	if !ok {
    56  		klog.Errorf("failed to convert %v to *v1.Pod", obj)
    57  		return
    58  	}
    59  
    60  	if err := cc.syncTask(task); err != nil {
    61  		klog.Errorf("Failed to sync pod <%v/%v>, retry it, err %v", task.Namespace, task.Name, err)
    62  		cc.resyncTask(task)
    63  	}
    64  }
    65  
    66  func (cc *jobcontroller) syncTask(oldTask *v1.Pod) error {
    67  	newPod, err := cc.kubeClient.CoreV1().Pods(oldTask.Namespace).Get(context.TODO(), oldTask.Name, metav1.GetOptions{})
    68  	if err != nil {
    69  		if errors.IsNotFound(err) {
    70  			if err := cc.cache.DeletePod(oldTask); err != nil {
    71  				klog.Errorf("failed to delete cache pod <%v/%v>, err %v.", oldTask.Namespace, oldTask.Name, err)
    72  				return err
    73  			}
    74  			klog.V(3).Infof("Pod <%v/%v> was deleted, removed from cache.", oldTask.Namespace, oldTask.Name)
    75  
    76  			return nil
    77  		}
    78  		return fmt.Errorf("failed to get Pod <%v/%v>: err %v", oldTask.Namespace, oldTask.Name, err)
    79  	}
    80  
    81  	return cc.cache.UpdatePod(newPod)
    82  }
    83  
    84  func (cc *jobcontroller) resyncTask(task *v1.Pod) {
    85  	cc.errTasks.AddRateLimited(task)
    86  }