k8s.io/kubernetes@v1.29.3/pkg/kubelet/checkpointmanager/checkpoint_manager.go (about)

     1  /*
     2  Copyright 2018 The Kubernetes 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 checkpointmanager
    18  
    19  import (
    20  	"fmt"
    21  	"sync"
    22  
    23  	"k8s.io/kubernetes/pkg/kubelet/checkpointmanager/errors"
    24  	utilstore "k8s.io/kubernetes/pkg/kubelet/util/store"
    25  	utilfs "k8s.io/kubernetes/pkg/util/filesystem"
    26  )
    27  
    28  // Checkpoint provides the process checkpoint data
    29  type Checkpoint interface {
    30  	MarshalCheckpoint() ([]byte, error)
    31  	UnmarshalCheckpoint(blob []byte) error
    32  	VerifyChecksum() error
    33  }
    34  
    35  // CheckpointManager provides the interface to manage checkpoint
    36  type CheckpointManager interface {
    37  	// CreateCheckpoint persists checkpoint in CheckpointStore. checkpointKey is the key for utilstore to locate checkpoint.
    38  	// For file backed utilstore, checkpointKey is the file name to write the checkpoint data.
    39  	CreateCheckpoint(checkpointKey string, checkpoint Checkpoint) error
    40  	// GetCheckpoint retrieves checkpoint from CheckpointStore.
    41  	GetCheckpoint(checkpointKey string, checkpoint Checkpoint) error
    42  	// WARNING: RemoveCheckpoint will not return error if checkpoint does not exist.
    43  	RemoveCheckpoint(checkpointKey string) error
    44  	// ListCheckpoint returns the list of existing checkpoints.
    45  	ListCheckpoints() ([]string, error)
    46  }
    47  
    48  // impl is an implementation of CheckpointManager. It persists checkpoint in CheckpointStore
    49  type impl struct {
    50  	path  string
    51  	store utilstore.Store
    52  	mutex sync.Mutex
    53  }
    54  
    55  // NewCheckpointManager returns a new instance of a checkpoint manager
    56  func NewCheckpointManager(checkpointDir string) (CheckpointManager, error) {
    57  	fstore, err := utilstore.NewFileStore(checkpointDir, &utilfs.DefaultFs{})
    58  	if err != nil {
    59  		return nil, err
    60  	}
    61  
    62  	return &impl{path: checkpointDir, store: fstore}, nil
    63  }
    64  
    65  // CreateCheckpoint persists checkpoint in CheckpointStore.
    66  func (manager *impl) CreateCheckpoint(checkpointKey string, checkpoint Checkpoint) error {
    67  	manager.mutex.Lock()
    68  	defer manager.mutex.Unlock()
    69  	blob, err := checkpoint.MarshalCheckpoint()
    70  	if err != nil {
    71  		return err
    72  	}
    73  	return manager.store.Write(checkpointKey, blob)
    74  }
    75  
    76  // GetCheckpoint retrieves checkpoint from CheckpointStore.
    77  func (manager *impl) GetCheckpoint(checkpointKey string, checkpoint Checkpoint) error {
    78  	manager.mutex.Lock()
    79  	defer manager.mutex.Unlock()
    80  	blob, err := manager.store.Read(checkpointKey)
    81  	if err != nil {
    82  		if err == utilstore.ErrKeyNotFound {
    83  			return errors.ErrCheckpointNotFound
    84  		}
    85  		return err
    86  	}
    87  	err = checkpoint.UnmarshalCheckpoint(blob)
    88  	if err == nil {
    89  		err = checkpoint.VerifyChecksum()
    90  	}
    91  	return err
    92  }
    93  
    94  // RemoveCheckpoint will not return error if checkpoint does not exist.
    95  func (manager *impl) RemoveCheckpoint(checkpointKey string) error {
    96  	manager.mutex.Lock()
    97  	defer manager.mutex.Unlock()
    98  	return manager.store.Delete(checkpointKey)
    99  }
   100  
   101  // ListCheckpoints returns the list of existing checkpoints.
   102  func (manager *impl) ListCheckpoints() ([]string, error) {
   103  	manager.mutex.Lock()
   104  	defer manager.mutex.Unlock()
   105  	keys, err := manager.store.List()
   106  	if err != nil {
   107  		return []string{}, fmt.Errorf("failed to list checkpoint store: %v", err)
   108  	}
   109  	return keys, nil
   110  }