|
| 1 | +/* |
| 2 | +Copyright 2020 The Kubernetes Authors. |
| 3 | +Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | +you may not use this file except in compliance with the License. |
| 5 | +You may obtain a copy of the License at |
| 6 | + http://www.apache.org/licenses/LICENSE-2.0 |
| 7 | +Unless required by applicable law or agreed to in writing, software |
| 8 | +distributed under the License is distributed on an "AS IS" BASIS, |
| 9 | +WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 10 | +See the License for the specific language governing permissions and |
| 11 | +limitations under the License. |
| 12 | +*/ |
| 13 | + |
| 14 | +// Package v2 is an out-of-tree only implementation of the AWS cloud provider. |
| 15 | +// It is not compatible with v1 and should only be used on new clusters. |
| 16 | +package v2 |
| 17 | + |
| 18 | +import ( |
| 19 | + "fmt" |
| 20 | + "strings" |
| 21 | + "time" |
| 22 | + |
| 23 | + "k8s.io/apimachinery/pkg/util/wait" |
| 24 | + "k8s.io/klog/v2" |
| 25 | + |
| 26 | + "github.com/aws/aws-sdk-go/aws" |
| 27 | + "github.com/aws/aws-sdk-go/service/ec2" |
| 28 | +) |
| 29 | + |
| 30 | +const ( |
| 31 | + // TagNameKubernetesClusterPrefix is the tag name we use to differentiate multiple |
| 32 | + // logically independent clusters running in the same AZ. |
| 33 | + // tag format: kubernetes.io/cluster/=<clusterName> |
| 34 | + TagNameKubernetesClusterPrefix = "kubernetes.io/cluster/" |
| 35 | + |
| 36 | + // createTag* is configuration of exponential backoff for CreateTag call. We |
| 37 | + // retry mainly because if we create an object, we cannot tag it until it is |
| 38 | + // "fully created" (eventual consistency). Starting with 1 second, doubling |
| 39 | + // it every step and taking 9 steps results in 255 second total waiting |
| 40 | + // time. |
| 41 | + createTagInitialDelay = 1 * time.Second |
| 42 | + createTagFactor = 2.0 |
| 43 | + createTagSteps = 9 |
| 44 | +) |
| 45 | + |
| 46 | +type awsTagging struct { |
| 47 | + // ClusterName is our cluster identifier: we tag AWS resources with this value, |
| 48 | + // and thus we can run two independent clusters in the same VPC or subnets. |
| 49 | + ClusterName string |
| 50 | +} |
| 51 | + |
| 52 | +// Extracts the cluster name from the given tags, if they are present |
| 53 | +// If duplicate tags are found, returns an error |
| 54 | +func findClusterName(tags []*ec2.Tag) (string, error) { |
| 55 | + clusterName := "" |
| 56 | + |
| 57 | + for _, tag := range tags { |
| 58 | + tagKey := aws.StringValue(tag.Key) |
| 59 | + if strings.HasPrefix(tagKey, TagNameKubernetesClusterPrefix) { |
| 60 | + name := aws.StringValue(tag.Value) |
| 61 | + if clusterName != "" { |
| 62 | + return "", fmt.Errorf("Found multiple cluster tags with prefix %s (%q and %q)", TagNameKubernetesClusterPrefix, clusterName, name) |
| 63 | + } |
| 64 | + clusterName = name |
| 65 | + } |
| 66 | + } |
| 67 | + |
| 68 | + return clusterName, nil |
| 69 | +} |
| 70 | + |
| 71 | +func (t *awsTagging) init(clusterName string) error { |
| 72 | + t.ClusterName = clusterName |
| 73 | + |
| 74 | + if clusterName != "" { |
| 75 | + klog.Infof("AWS cloud filtering on ClusterName: %v", clusterName) |
| 76 | + } else { |
| 77 | + return fmt.Errorf("AWS cloud failed to find ClusterName") |
| 78 | + } |
| 79 | + |
| 80 | + return nil |
| 81 | +} |
| 82 | + |
| 83 | +// Extracts a cluster name from the given tags, if one is present |
| 84 | +// If no clusterName is found, returns "", nil |
| 85 | +// If multiple (different) clusterNames are found, returns an error |
| 86 | +func (t *awsTagging) initFromTags(tags []*ec2.Tag) error { |
| 87 | + clusterName, err := findClusterName(tags) |
| 88 | + if err != nil { |
| 89 | + return err |
| 90 | + } |
| 91 | + |
| 92 | + if clusterName == "" { |
| 93 | + klog.Errorf("Tag %q not found; Kubernetes may behave unexpectedly.", TagNameKubernetesClusterPrefix) |
| 94 | + } |
| 95 | + |
| 96 | + return t.init(clusterName) |
| 97 | +} |
| 98 | + |
| 99 | +func (t *awsTagging) hasClusterTag(tags []*ec2.Tag) bool { |
| 100 | + // if the clusterName is not configured -- we consider all instances. |
| 101 | + if len(t.ClusterName) == 0 { |
| 102 | + return true |
| 103 | + } |
| 104 | + |
| 105 | + for _, tag := range tags { |
| 106 | + tagKey := aws.StringValue(tag.Key) |
| 107 | + if (tagKey == TagNameKubernetesClusterPrefix) && (aws.StringValue(tag.Value) == t.ClusterName) { |
| 108 | + return true |
| 109 | + } |
| 110 | + } |
| 111 | + |
| 112 | + return false |
| 113 | +} |
| 114 | + |
| 115 | +func (t *awsTagging) buildTags(additionalTags map[string]string, lifecycle string) map[string]string { |
| 116 | + tags := make(map[string]string) |
| 117 | + for k, v := range additionalTags { |
| 118 | + tags[k] = v |
| 119 | + } |
| 120 | + |
| 121 | + // no clusterName is a sign of misconfigured cluster, but we can't be tagging the resources with empty |
| 122 | + // strings |
| 123 | + if len(t.ClusterName) == 0 { |
| 124 | + return tags |
| 125 | + } |
| 126 | + |
| 127 | + // tag format: kubernetes.io/cluster/=<clusterName> |
| 128 | + tags[TagNameKubernetesClusterPrefix] = t.ClusterName |
| 129 | + |
| 130 | + return tags |
| 131 | +} |
| 132 | + |
| 133 | +// createTags calls EC2 CreateTags, but adds retry-on-failure logic |
| 134 | +// We retry mainly because if we create an object, we cannot tag it until it is "fully created" (eventual consistency) |
| 135 | +// The error code varies though (depending on what we are tagging), so we simply retry on all errors |
| 136 | +func (t *awsTagging) createTags(ec2Client EC2, resourceID string, lifecycle string, additionalTags map[string]string) error { |
| 137 | + tags := t.buildTags(additionalTags, lifecycle) |
| 138 | + |
| 139 | + if tags == nil || len(tags) == 0 { |
| 140 | + return nil |
| 141 | + } |
| 142 | + |
| 143 | + var awsTags []*ec2.Tag |
| 144 | + for k, v := range tags { |
| 145 | + tag := &ec2.Tag{ |
| 146 | + Key: aws.String(k), |
| 147 | + Value: aws.String(v), |
| 148 | + } |
| 149 | + awsTags = append(awsTags, tag) |
| 150 | + } |
| 151 | + |
| 152 | + backoff := wait.Backoff{ |
| 153 | + Duration: createTagInitialDelay, |
| 154 | + Factor: createTagFactor, |
| 155 | + Steps: createTagSteps, |
| 156 | + } |
| 157 | + |
| 158 | + request := &ec2.CreateTagsInput{ |
| 159 | + Resources: []*string{&resourceID}, |
| 160 | + Tags: awsTags, |
| 161 | + } |
| 162 | + |
| 163 | + var lastErr error |
| 164 | + err := wait.ExponentialBackoff(backoff, func() (bool, error) { |
| 165 | + _, err := ec2Client.CreateTags(request) |
| 166 | + if err == nil { |
| 167 | + return true, nil |
| 168 | + } |
| 169 | + |
| 170 | + // We could check that the error is retryable, but the error code changes based on what we are tagging |
| 171 | + // SecurityGroup: InvalidGroup.NotFound |
| 172 | + klog.V(2).Infof("Failed to create tags; will retry. Error was %q", err) |
| 173 | + lastErr = err |
| 174 | + return false, nil |
| 175 | + }) |
| 176 | + if err == wait.ErrWaitTimeout { |
| 177 | + // return real CreateTags error instead of timeout |
| 178 | + err = lastErr |
| 179 | + } |
| 180 | + |
| 181 | + return err |
| 182 | +} |
| 183 | + |
| 184 | +func (t *awsTagging) clusterName() string { |
| 185 | + return t.ClusterName |
| 186 | +} |
0 commit comments