forked from moby/swarmkit
/
cluster.go
309 lines (272 loc) · 10.1 KB
/
cluster.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
package controlapi
import (
"context"
"strings"
"time"
"github.com/docker/swarmkit/api"
"github.com/docker/swarmkit/ca"
"github.com/docker/swarmkit/log"
"github.com/docker/swarmkit/manager/encryption"
"github.com/docker/swarmkit/manager/state/store"
gogotypes "github.com/gogo/protobuf/types"
"google.golang.org/grpc/codes"
"google.golang.org/grpc/status"
)
const (
// expiredCertGrace is the amount of time to keep a node in the
// blacklist beyond its certificate expiration timestamp.
expiredCertGrace = 24 * time.Hour * 7
// inbuilt default subnet size
inbuiltSubnetSize = 24
// VXLAN default port
defaultVXLANPort = 4789
)
var (
// inbuilt default address pool
inbuiltDefaultAddressPool = []string{"10.0.0.0/8"}
)
func validateClusterSpec(spec *api.ClusterSpec) error {
if spec == nil {
return status.Errorf(codes.InvalidArgument, errInvalidArgument.Error())
}
// Validate that expiry time being provided is valid, and over our minimum
if spec.CAConfig.NodeCertExpiry != nil {
expiry, err := gogotypes.DurationFromProto(spec.CAConfig.NodeCertExpiry)
if err != nil {
return status.Errorf(codes.InvalidArgument, errInvalidArgument.Error())
}
if expiry < ca.MinNodeCertExpiration {
return status.Errorf(codes.InvalidArgument, "minimum certificate expiry time is: %s", ca.MinNodeCertExpiration)
}
}
// Validate that AcceptancePolicies only include Secrets that are bcrypted
// TODO(diogo): Add a global list of acceptance algorithms. We only support bcrypt for now.
if len(spec.AcceptancePolicy.Policies) > 0 {
for _, policy := range spec.AcceptancePolicy.Policies {
if policy.Secret != nil && strings.ToLower(policy.Secret.Alg) != "bcrypt" {
return status.Errorf(codes.InvalidArgument, "hashing algorithm is not supported: %s", policy.Secret.Alg)
}
}
}
// Validate that heartbeatPeriod time being provided is valid
if spec.Dispatcher.HeartbeatPeriod != nil {
heartbeatPeriod, err := gogotypes.DurationFromProto(spec.Dispatcher.HeartbeatPeriod)
if err != nil {
return status.Errorf(codes.InvalidArgument, errInvalidArgument.Error())
}
if heartbeatPeriod < 0 {
return status.Errorf(codes.InvalidArgument, "heartbeat time period cannot be a negative duration")
}
}
if spec.Annotations.Name != store.DefaultClusterName {
return status.Errorf(codes.InvalidArgument, "modification of cluster name is not allowed")
}
return nil
}
// GetCluster returns a Cluster given a ClusterID.
// - Returns `InvalidArgument` if ClusterID is not provided.
// - Returns `NotFound` if the Cluster is not found.
func (s *Server) GetCluster(ctx context.Context, request *api.GetClusterRequest) (*api.GetClusterResponse, error) {
if request.ClusterID == "" {
return nil, status.Errorf(codes.InvalidArgument, errInvalidArgument.Error())
}
var cluster *api.Cluster
s.store.View(func(tx store.ReadTx) {
cluster = store.GetCluster(tx, request.ClusterID)
})
if cluster == nil {
return nil, status.Errorf(codes.NotFound, "cluster %s not found", request.ClusterID)
}
redactedClusters := redactClusters([]*api.Cluster{cluster})
// WARN: we should never return cluster here. We need to redact the private fields first.
return &api.GetClusterResponse{
Cluster: redactedClusters[0],
}, nil
}
// UpdateCluster updates a Cluster referenced by ClusterID with the given ClusterSpec.
// - Returns `NotFound` if the Cluster is not found.
// - Returns `InvalidArgument` if the ClusterSpec is malformed.
// - Returns `Unimplemented` if the ClusterSpec references unimplemented features.
// - Returns an error if the update fails.
func (s *Server) UpdateCluster(ctx context.Context, request *api.UpdateClusterRequest) (*api.UpdateClusterResponse, error) {
if request.ClusterID == "" || request.ClusterVersion == nil {
return nil, status.Errorf(codes.InvalidArgument, errInvalidArgument.Error())
}
if err := validateClusterSpec(request.Spec); err != nil {
return nil, err
}
var cluster *api.Cluster
err := s.store.Update(func(tx store.Tx) error {
cluster = store.GetCluster(tx, request.ClusterID)
if cluster == nil {
return status.Errorf(codes.NotFound, "cluster %s not found", request.ClusterID)
}
// This ensures that we have the current rootCA with which to generate tokens (expiration doesn't matter
// for generating the tokens)
rootCA, err := ca.RootCAFromAPI(ctx, &cluster.RootCA, ca.DefaultNodeCertExpiration)
if err != nil {
log.G(ctx).WithField(
"method", "(*controlapi.Server).UpdateCluster").WithError(err).Error("invalid cluster root CA")
return status.Errorf(codes.Internal, "error loading cluster rootCA for update")
}
cluster.Meta.Version = *request.ClusterVersion
cluster.Spec = *request.Spec.Copy()
expireBlacklistedCerts(cluster)
if request.Rotation.WorkerJoinToken {
cluster.RootCA.JoinTokens.Worker = ca.GenerateJoinToken(&rootCA, cluster.FIPS)
}
if request.Rotation.ManagerJoinToken {
cluster.RootCA.JoinTokens.Manager = ca.GenerateJoinToken(&rootCA, cluster.FIPS)
}
updatedRootCA, err := validateCAConfig(ctx, s.securityConfig, cluster)
if err != nil {
return err
}
cluster.RootCA = *updatedRootCA
var unlockKeys []*api.EncryptionKey
var managerKey *api.EncryptionKey
for _, eKey := range cluster.UnlockKeys {
if eKey.Subsystem == ca.ManagerRole {
if !cluster.Spec.EncryptionConfig.AutoLockManagers {
continue
}
managerKey = eKey
}
unlockKeys = append(unlockKeys, eKey)
}
switch {
case !cluster.Spec.EncryptionConfig.AutoLockManagers:
break
case managerKey == nil:
unlockKeys = append(unlockKeys, &api.EncryptionKey{
Subsystem: ca.ManagerRole,
Key: encryption.GenerateSecretKey(),
})
case request.Rotation.ManagerUnlockKey:
managerKey.Key = encryption.GenerateSecretKey()
}
cluster.UnlockKeys = unlockKeys
return store.UpdateCluster(tx, cluster)
})
if err != nil {
return nil, err
}
redactedClusters := redactClusters([]*api.Cluster{cluster})
// WARN: we should never return cluster here. We need to redact the private fields first.
return &api.UpdateClusterResponse{
Cluster: redactedClusters[0],
}, nil
}
func filterClusters(candidates []*api.Cluster, filters ...func(*api.Cluster) bool) []*api.Cluster {
result := []*api.Cluster{}
for _, c := range candidates {
match := true
for _, f := range filters {
if !f(c) {
match = false
break
}
}
if match {
result = append(result, c)
}
}
return result
}
// ListClusters returns a list of all clusters.
func (s *Server) ListClusters(ctx context.Context, request *api.ListClustersRequest) (*api.ListClustersResponse, error) {
var (
clusters []*api.Cluster
err error
)
s.store.View(func(tx store.ReadTx) {
switch {
case request.Filters != nil && len(request.Filters.Names) > 0:
clusters, err = store.FindClusters(tx, buildFilters(store.ByName, request.Filters.Names))
case request.Filters != nil && len(request.Filters.NamePrefixes) > 0:
clusters, err = store.FindClusters(tx, buildFilters(store.ByNamePrefix, request.Filters.NamePrefixes))
case request.Filters != nil && len(request.Filters.IDPrefixes) > 0:
clusters, err = store.FindClusters(tx, buildFilters(store.ByIDPrefix, request.Filters.IDPrefixes))
default:
clusters, err = store.FindClusters(tx, store.All)
}
})
if err != nil {
return nil, err
}
if request.Filters != nil {
clusters = filterClusters(clusters,
func(e *api.Cluster) bool {
return filterContains(e.Spec.Annotations.Name, request.Filters.Names)
},
func(e *api.Cluster) bool {
return filterContainsPrefix(e.Spec.Annotations.Name, request.Filters.NamePrefixes)
},
func(e *api.Cluster) bool {
return filterContainsPrefix(e.ID, request.Filters.IDPrefixes)
},
func(e *api.Cluster) bool {
return filterMatchLabels(e.Spec.Annotations.Labels, request.Filters.Labels)
},
)
}
// WARN: we should never return cluster here. We need to redact the private fields first.
return &api.ListClustersResponse{
Clusters: redactClusters(clusters),
}, nil
}
// redactClusters is a method that enforces a whitelist of fields that are ok to be
// returned in the Cluster object. It should filter out all sensitive information.
func redactClusters(clusters []*api.Cluster) []*api.Cluster {
var redactedClusters []*api.Cluster
// Only add public fields to the new clusters
for _, cluster := range clusters {
// Copy all the mandatory fields
// Do not copy secret keys
redactedSpec := cluster.Spec.Copy()
redactedSpec.CAConfig.SigningCAKey = nil
// the cert is not a secret, but if API users get the cluster spec and then update,
// then because the cert is included but not the key, the user can get update errors
// or unintended consequences (such as telling swarm to forget about the key so long
// as there is a corresponding external CA)
redactedSpec.CAConfig.SigningCACert = nil
redactedRootCA := cluster.RootCA.Copy()
redactedRootCA.CAKey = nil
if r := redactedRootCA.RootRotation; r != nil {
r.CAKey = nil
}
newCluster := &api.Cluster{
ID: cluster.ID,
Meta: cluster.Meta,
Spec: *redactedSpec,
RootCA: *redactedRootCA,
BlacklistedCertificates: cluster.BlacklistedCertificates,
DefaultAddressPool: cluster.DefaultAddressPool,
SubnetSize: cluster.SubnetSize,
VXLANUDPPort: cluster.VXLANUDPPort,
}
if newCluster.DefaultAddressPool == nil {
// This is just for CLI display. Set the inbuilt default pool for
// user reference.
newCluster.DefaultAddressPool = inbuiltDefaultAddressPool
newCluster.SubnetSize = inbuiltSubnetSize
}
if newCluster.VXLANUDPPort == 0 {
newCluster.VXLANUDPPort = defaultVXLANPort
}
redactedClusters = append(redactedClusters, newCluster)
}
return redactedClusters
}
func expireBlacklistedCerts(cluster *api.Cluster) {
nowMinusGrace := time.Now().Add(-expiredCertGrace)
for cn, blacklistedCert := range cluster.BlacklistedCertificates {
if blacklistedCert.Expiry == nil {
continue
}
expiry, err := gogotypes.TimestampFromProto(blacklistedCert.Expiry)
if err == nil && nowMinusGrace.After(expiry) {
delete(cluster.BlacklistedCertificates, cn)
}
}
}