sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 1 | /* |
| 2 | * Copyright 2018-present Open Networking Foundation |
| 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 main |
| 18 | |
| 19 | import ( |
| 20 | //"os" |
| 21 | "fmt" |
| 22 | "time" |
| 23 | "regexp" |
| 24 | "errors" |
| 25 | "strconv" |
| 26 | //"io/ioutil" |
| 27 | //"encoding/json" |
| 28 | |
| 29 | "k8s.io/client-go/rest" |
| 30 | "google.golang.org/grpc" |
| 31 | "golang.org/x/net/context" |
| 32 | "k8s.io/client-go/kubernetes" |
| 33 | "github.com/golang/protobuf/ptypes" |
| 34 | //"k8s.io/apimachinery/pkg/api/errors" |
| 35 | "github.com/opencord/voltha-go/common/log" |
| 36 | kafka "github.com/opencord/voltha-go/kafka" |
| 37 | metav1 "k8s.io/apimachinery/pkg/apis/meta/v1" |
| 38 | empty "github.com/golang/protobuf/ptypes/empty" |
| 39 | vpb "github.com/opencord/voltha-go/protos/voltha" |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 40 | cmn "github.com/opencord/voltha-go/protos/common" |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 41 | pb "github.com/opencord/voltha-go/protos/afrouter" |
| 42 | ic "github.com/opencord/voltha-go/protos/inter_container" |
| 43 | ) |
| 44 | |
| 45 | type configConn struct { |
| 46 | Server string `json:"Server"` |
| 47 | Cluster string `json:"Cluster"` |
| 48 | Backend string `json:"Backend"` |
| 49 | connections map[string]connection |
| 50 | } |
| 51 | |
| 52 | type connection struct { |
| 53 | Name string `json:"Connection"` |
| 54 | Addr string `json:"Addr"` |
| 55 | Port uint64 `json:"Port"` |
| 56 | } |
| 57 | |
| 58 | type rwPod struct { |
| 59 | name string |
| 60 | ipAddr string |
| 61 | node string |
| 62 | devIds map[string]struct{} |
| 63 | backend string |
| 64 | connection string |
| 65 | } |
| 66 | |
| 67 | type podTrack struct { |
| 68 | pod *rwPod |
| 69 | dn bool |
| 70 | } |
| 71 | |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 72 | var nPods int = 6 |
| 73 | |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 74 | // Topic is affinityRouter |
| 75 | // port: 9092 |
| 76 | |
| 77 | func newKafkaClient(clientType string, host string, port int, instanceID string) (kafka.Client, error) { |
| 78 | |
| 79 | log.Infow("kafka-client-type", log.Fields{"client": clientType}) |
| 80 | switch clientType { |
| 81 | case "sarama": |
| 82 | return kafka.NewSaramaClient( |
| 83 | kafka.Host(host), |
| 84 | kafka.Port(port), |
| 85 | kafka.ConsumerType(kafka.GroupCustomer), |
| 86 | kafka.ProducerReturnOnErrors(true), |
| 87 | kafka.ProducerReturnOnSuccess(true), |
| 88 | kafka.ProducerMaxRetries(6), |
| 89 | kafka.NumPartitions(3), |
| 90 | kafka.ConsumerGroupName(instanceID), |
| 91 | kafka.ConsumerGroupPrefix(instanceID), |
| 92 | kafka.AutoCreateTopic(false), |
| 93 | kafka.ProducerFlushFrequency(5), |
| 94 | kafka.ProducerRetryBackoff(time.Millisecond*30)), nil |
| 95 | } |
| 96 | return nil, errors.New("unsupported-client-type") |
| 97 | } |
| 98 | |
| 99 | |
| 100 | func k8sClientSet() *kubernetes.Clientset { |
| 101 | // creates the in-cluster config |
| 102 | config, err := rest.InClusterConfig() |
| 103 | if err != nil { |
| 104 | panic(err.Error()) |
| 105 | } |
| 106 | // creates the clientset |
| 107 | clientset, err := kubernetes.NewForConfig(config) |
| 108 | if err != nil { |
| 109 | panic(err.Error()) |
| 110 | } |
| 111 | |
| 112 | return clientset |
| 113 | } |
| 114 | |
| 115 | |
| 116 | func connect(addr string) (*grpc.ClientConn, error) { |
| 117 | for ctr :=0 ; ctr < 100; ctr++ { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 118 | log.Debugf("Trying to connect to %s", addr) |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 119 | conn, err := grpc.Dial(addr, grpc.WithInsecure()) |
| 120 | if err != nil { |
| 121 | log.Debugf("Attempt to connect failed, retrying %v:", err) |
| 122 | } else { |
| 123 | log.Debugf("Connection succeeded") |
| 124 | return conn,err |
| 125 | } |
| 126 | time.Sleep(10 * time.Second) |
| 127 | } |
| 128 | log.Debugf("Too many connection attempts, giving up!") |
| 129 | return nil,errors.New("Timeout attempting to conect") |
| 130 | } |
| 131 | |
| 132 | func getRwPods(cs *kubernetes.Clientset, coreFilter * regexp.Regexp) []*rwPod { |
| 133 | var rtrn []*rwPod |
| 134 | |
| 135 | pods, err := cs.CoreV1().Pods("").List(metav1.ListOptions{}) |
| 136 | if err != nil { |
| 137 | panic(err.Error()) |
| 138 | } |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 139 | //log.Debugf("There are a total of %d pods in the cluster\n", len(pods.Items)) |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 140 | |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 141 | for _,v := range pods.Items { |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 142 | if v.Namespace == "voltha" && coreFilter.MatchString(v.Name) { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 143 | log.Debugf("Namespace: %s, PodName: %s, PodIP: %s, Host: %s\n", v.Namespace, v.Name, |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 144 | v.Status.PodIP, v.Spec.NodeName) |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 145 | // Only add the pod if it has an IP address. If it doesn't then it likely crashed and |
| 146 | // and is still in the process of getting re-started. |
| 147 | if v.Status.PodIP != "" { |
| 148 | rtrn = append(rtrn, &rwPod{name:v.Name,ipAddr:v.Status.PodIP,node:v.Spec.NodeName, |
| 149 | devIds:make(map[string]struct{}), backend:"", connection:""}) |
| 150 | } |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 151 | } |
| 152 | } |
| 153 | return rtrn |
| 154 | } |
| 155 | |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 156 | func reconcilePodDeviceIds(pod * rwPod, ids map[string]struct{}) bool { |
| 157 | var idList cmn.IDs |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 158 | for k,_ := range ids { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 159 | idList.Items = append(idList.Items, &cmn.ID{Id:k}) |
| 160 | } |
| 161 | conn,err := connect(pod.ipAddr+":50057") |
| 162 | defer conn.Close() |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 163 | if err != nil { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 164 | log.Debugf("Could not query devices from %s, could not connect", pod.name) |
| 165 | return false |
| 166 | } |
| 167 | client := vpb.NewVolthaServiceClient(conn) |
| 168 | _,err = client.ReconcileDevices(context.Background(), &idList) |
| 169 | if err != nil { |
| 170 | log.Error(err) |
| 171 | return false |
| 172 | } |
| 173 | |
| 174 | return true |
| 175 | } |
| 176 | |
| 177 | func queryPodDeviceIds(pod * rwPod) map[string]struct{} { |
| 178 | var rtrn map[string]struct{} = make(map[string]struct{}) |
| 179 | // Open a connection to the pod |
| 180 | // port 50057 |
| 181 | conn, err := connect(pod.ipAddr+":50057") |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 182 | if err != nil { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 183 | log.Debugf("Could not query devices from %s, could not connect", pod.name) |
| 184 | return rtrn |
| 185 | } |
| 186 | defer conn.Close() |
| 187 | client := vpb.NewVolthaServiceClient(conn) |
| 188 | devs,err := client.ListDeviceIds(context.Background(), &empty.Empty{}) |
| 189 | if err != nil { |
| 190 | log.Error(err) |
| 191 | return rtrn |
| 192 | } |
| 193 | for _,dv := range devs.Items { |
| 194 | rtrn[dv.Id]=struct{}{} |
| 195 | } |
| 196 | |
| 197 | return rtrn |
| 198 | } |
| 199 | |
| 200 | func queryDeviceIds(pods []*rwPod) { |
| 201 | for pk,_ := range pods { |
| 202 | // Keep the old Id list if a new list is not returned |
| 203 | if idList := queryPodDeviceIds(pods[pk]); len(idList) != 0 { |
| 204 | pods[pk].devIds = idList |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 205 | } |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 206 | } |
| 207 | } |
| 208 | |
| 209 | func allEmpty(pods []*rwPod) bool { |
| 210 | for k,_ := range pods { |
| 211 | if len(pods[k].devIds) != 0 { |
| 212 | return false |
| 213 | } |
| 214 | } |
| 215 | return true |
| 216 | } |
| 217 | |
| 218 | //func groupEmptyCores(pods []*rwPod) [][]*rwPod { |
| 219 | // return [][]*rwPod{} |
| 220 | //} |
| 221 | |
| 222 | //func groupPods(pods []*rwPod) [][]*rwPod { |
| 223 | |
| 224 | // if allEmpty(pods) == true { |
| 225 | // return groupEmptyCores(pods) |
| 226 | // } else { |
| 227 | // return groupPopulatedCores(pods) |
| 228 | // } |
| 229 | //} |
| 230 | |
| 231 | func rmPod(pods []*rwPod, idx int) []*rwPod { |
| 232 | return append(pods[:idx],pods[idx+1:]...) |
| 233 | } |
| 234 | |
| 235 | func groupIntersectingPods1(pods []*rwPod, podCt int) ([][]*rwPod,[]*rwPod) { |
| 236 | var rtrn [][]*rwPod |
| 237 | var out []*rwPod |
| 238 | |
| 239 | for { |
| 240 | if len(pods) == 0 { |
| 241 | break |
| 242 | } |
| 243 | if len(pods[0].devIds) == 0 { // Ignore pods with no devices |
| 244 | ////log.Debugf("%s empty pod", pd[k].pod.name) |
| 245 | out = append(out, pods[0]) |
| 246 | pods = rmPod(pods, 0) |
| 247 | continue |
| 248 | } |
| 249 | // Start a pod group with this pod |
| 250 | var grp []*rwPod |
| 251 | grp = append(grp, pods[0]) |
| 252 | pods = rmPod(pods,0) |
| 253 | //log.Debugf("Creating new group %s", pd[k].pod.name) |
| 254 | // Find the peer pod based on device overlap |
| 255 | // It's ok if one isn't found, an empty one will be used instead |
| 256 | for k,_ := range pods { |
| 257 | if len(pods[k].devIds) == 0 { // Skip pods with no devices |
| 258 | //log.Debugf("%s empty pod", pd[k1].pod.name) |
| 259 | continue |
| 260 | } |
| 261 | if intersect(grp[0].devIds, pods[k].devIds) == true { |
| 262 | //log.Debugf("intersection found %s:%s", pd[k].pod.name, pd[k1].pod.name) |
| 263 | if grp[0].node == pods[k].node { |
| 264 | // This should never happen |
| 265 | log.Errorf("pods %s and %s intersect and are on the same server!! Not pairing", |
| 266 | grp[0].name, pods[k].name) |
| 267 | continue |
| 268 | } |
| 269 | grp = append(grp, pods[k]) |
| 270 | pods = rmPod(pods, k) |
| 271 | break |
| 272 | |
| 273 | } |
| 274 | } |
| 275 | rtrn = append(rtrn, grp) |
| 276 | //log.Debugf("Added group %s", grp[0].name) |
| 277 | // Check if the number of groups = half the pods, if so all groups are started. |
| 278 | if len(rtrn) == podCt >> 1 { |
| 279 | // Append any remaining pods to out |
| 280 | out = append(out,pods[0:]...) |
| 281 | break |
| 282 | } |
| 283 | } |
| 284 | return rtrn,out |
| 285 | } |
| 286 | |
| 287 | func groupIntersectingPods(pd []*podTrack) ([][]*rwPod,[]*podTrack) { |
| 288 | var rtrn [][]*rwPod |
| 289 | |
| 290 | for k,_ := range pd { |
| 291 | if pd[k].dn == true { // Already processed? |
| 292 | //log.Debugf("%s already processed", pd[k].pod.name) |
| 293 | continue |
| 294 | } |
| 295 | if len(pd[k].pod.devIds) == 0 { // Ignore pods with no devices |
| 296 | ////log.Debugf("%s empty pod", pd[k].pod.name) |
| 297 | continue |
| 298 | } |
| 299 | // Start a pod group with this pod |
| 300 | var grp []*rwPod |
| 301 | grp = append(grp, pd[k].pod) |
| 302 | pd[k].dn = true |
| 303 | //log.Debugf("Creating new group %s", pd[k].pod.name) |
| 304 | // Find the peer pod based on device overlap |
| 305 | // It's ok if one isn't found, an empty one will be used instead |
| 306 | for k1,_ := range pd { |
| 307 | if pd[k1].dn == true { // Skip over eliminated pods |
| 308 | //log.Debugf("%s eliminated pod", pd[k1].pod.name) |
| 309 | continue |
| 310 | } |
| 311 | if len(pd[k1].pod.devIds) == 0 { // Skip pods with no devices |
| 312 | //log.Debugf("%s empty pod", pd[k1].pod.name) |
| 313 | continue |
| 314 | } |
| 315 | if intersect(pd[k].pod.devIds, pd[k1].pod.devIds) == true { |
| 316 | //log.Debugf("intersection found %s:%s", pd[k].pod.name, pd[k1].pod.name) |
| 317 | if pd[k].pod.node == pd[k1].pod.node { |
| 318 | // This should never happen |
| 319 | log.Errorf("pods %s and %s intersect and are on the same server!! Not pairing", |
| 320 | pd[k].pod.name, pd[k1].pod.name) |
| 321 | continue |
| 322 | } |
| 323 | pd[k1].dn = true |
| 324 | grp = append(grp, pd[k1].pod) |
| 325 | break |
| 326 | } |
| 327 | } |
| 328 | rtrn = append(rtrn, grp) |
| 329 | //log.Debugf("Added group %s", grp[0].name) |
| 330 | // Check if the number of groups = half the pods, if so all groups are started. |
| 331 | if len(rtrn) == len(pd) >> 1 { |
| 332 | break |
| 333 | } |
| 334 | } |
| 335 | return rtrn,pd |
| 336 | } |
| 337 | |
| 338 | func unallocPodCount(pd []*podTrack) int { |
| 339 | var rtrn int = 0 |
| 340 | for _,v := range pd { |
| 341 | if v.dn == false { |
| 342 | rtrn++ |
| 343 | } |
| 344 | } |
| 345 | return rtrn |
| 346 | } |
| 347 | |
| 348 | |
| 349 | func sameNode(pod *rwPod, grps [][]*rwPod) bool { |
| 350 | for _,v := range grps { |
| 351 | if v[0].node == pod.node { |
| 352 | return true |
| 353 | } |
| 354 | if len(v) == 2 && v[1].node == pod.node { |
| 355 | return true |
| 356 | } |
| 357 | } |
| 358 | return false |
| 359 | } |
| 360 | |
| 361 | func startRemainingGroups1(grps [][]*rwPod, pods []*rwPod, podCt int) ([][]*rwPod, []*rwPod) { |
| 362 | var grp []*rwPod |
| 363 | |
| 364 | for k,_ := range pods { |
| 365 | if sameNode(pods[k], grps) { |
| 366 | continue |
| 367 | } |
| 368 | grp = []*rwPod{} |
| 369 | grp = append(grp, pods[k]) |
| 370 | pods = rmPod(pods, k) |
| 371 | grps = append(grps, grp) |
| 372 | if len(grps) == podCt >> 1 { |
| 373 | break |
| 374 | } |
| 375 | } |
| 376 | return grps, pods |
| 377 | } |
| 378 | |
| 379 | func startRemainingGroups(grps [][]*rwPod, pd []*podTrack) ([][]*rwPod, []*podTrack) { |
| 380 | var grp []*rwPod |
| 381 | |
| 382 | for k,_ := range pd { |
| 383 | if sameNode(pd[k].pod, grps) == true { |
| 384 | continue |
| 385 | } |
| 386 | grp = append(grp, pd[k].pod) |
| 387 | grps = append(grps, grp) |
| 388 | pd[k].dn = true |
| 389 | if len(grps) == len(pd) >> 1 { |
| 390 | break |
| 391 | } |
| 392 | } |
| 393 | return grps, pd |
| 394 | } |
| 395 | |
| 396 | func hasSingleSecondNode(grp []*rwPod) bool { |
| 397 | var srvrs map[string]struct{} = make(map[string]struct{}) |
| 398 | for k,_ := range grp { |
| 399 | if k == 0 { |
| 400 | continue // Ignore the first item |
| 401 | } |
| 402 | srvrs[grp[k].node] = struct{}{} |
| 403 | } |
| 404 | if len(srvrs) == 1 { |
| 405 | return true |
| 406 | } |
| 407 | return false |
| 408 | } |
| 409 | |
| 410 | func addNode(grps [][]*rwPod, idx *rwPod, item *rwPod) [][]*rwPod { |
| 411 | for k,_ := range grps { |
| 412 | if grps[k][0].name == idx.name { |
| 413 | grps[k] = append(grps[k], item) |
| 414 | return grps |
| 415 | } |
| 416 | } |
| 417 | // TODO: Error checking required here. |
| 418 | return grps |
| 419 | } |
| 420 | |
| 421 | func removeNode(grps [][]*rwPod, item *rwPod) [][]*rwPod { |
| 422 | for k,_ := range grps { |
| 423 | for k1,_ := range grps[k] { |
| 424 | if grps[k][k1].name == item.name { |
| 425 | grps[k] = append(grps[k][:k1],grps[k][k1+1:]...) |
| 426 | break |
| 427 | } |
| 428 | } |
| 429 | } |
| 430 | return grps |
| 431 | } |
| 432 | |
| 433 | func groupRemainingPods1(grps [][]*rwPod, pods []*rwPod) [][]*rwPod { |
| 434 | var lgrps [][]*rwPod |
| 435 | // All groups must be started when this function is called. |
| 436 | // Copy incomplete groups |
| 437 | for k,_ := range grps { |
| 438 | if len(grps[k]) != 2 { |
| 439 | lgrps = append(lgrps, grps[k]) |
| 440 | } |
| 441 | } |
| 442 | |
| 443 | // Add all pairing candidates to each started group. |
| 444 | for k,_ := range pods { |
| 445 | for k2,_ := range lgrps { |
| 446 | if lgrps[k2][0].node != pods[k].node { |
| 447 | lgrps[k2] = append(lgrps[k2], pods[k]) |
| 448 | } |
| 449 | } |
| 450 | } |
| 451 | |
| 452 | //TODO: If any member of lgrps doesn't have at least 2 |
| 453 | // nodes something is wrong. Check for that here |
| 454 | |
| 455 | for { |
| 456 | for { // Address groups with only a single server choice |
| 457 | var ssn bool = false |
| 458 | |
| 459 | for k,_ := range lgrps { |
| 460 | // Now if any of the groups only have a single |
| 461 | // node as the choice for the second member |
| 462 | // address that one first. |
| 463 | if hasSingleSecondNode(lgrps[k]) == true { |
| 464 | ssn = true |
| 465 | // Add this pairing to the groups |
| 466 | grps = addNode(grps, lgrps[k][0], lgrps[k][1]) |
| 467 | // Since this node is now used, remove it from all |
| 468 | // remaining tenative groups |
| 469 | lgrps = removeNode(lgrps, lgrps[k][1]) |
| 470 | // Now remove this group completely since |
| 471 | // it's been addressed |
| 472 | lgrps = append(lgrps[:k],lgrps[k+1:]...) |
| 473 | break |
| 474 | } |
| 475 | } |
| 476 | if ssn == false { |
| 477 | break |
| 478 | } |
| 479 | } |
| 480 | // Now adress one of the remaining groups |
| 481 | if len(lgrps) == 0 { |
| 482 | break // Nothing left to do, exit the loop |
| 483 | } |
| 484 | grps = addNode(grps, lgrps[0][0], lgrps[0][1]) |
| 485 | lgrps = removeNode(lgrps, lgrps[0][1]) |
| 486 | lgrps = append(lgrps[:0],lgrps[1:]...) |
| 487 | } |
| 488 | return grps |
| 489 | } |
| 490 | |
| 491 | func groupRemainingPods(grps [][]*rwPod, pd []*podTrack) [][]*rwPod{ |
| 492 | var lgrps [][]*rwPod |
| 493 | // All groups must be started when this function is called. |
| 494 | // Copy incomplete groups |
| 495 | for k,_ := range grps { |
| 496 | if len(grps[k]) != 2 { |
| 497 | lgrps = append(lgrps, grps[k]) |
| 498 | } |
| 499 | } |
| 500 | |
| 501 | // Add all pairing candidates to each started group. |
| 502 | for k,_ := range pd { |
| 503 | if pd[k].dn == true { |
| 504 | continue |
| 505 | } |
| 506 | for k2,_ := range lgrps { |
| 507 | if lgrps[k2][0].node != pd[k].pod.node { |
| 508 | lgrps[k2] = append(lgrps[k2], pd[k].pod) |
| 509 | } |
| 510 | } |
| 511 | } |
| 512 | |
| 513 | //TODO: If any member of lgrps doesn't have at least 2 |
| 514 | // nodes something is wrong. Check for that here |
| 515 | |
| 516 | for { |
| 517 | for { // Address groups with only a single server choice |
| 518 | var ssn bool = false |
| 519 | |
| 520 | for k,_ := range lgrps { |
| 521 | // Now if any of the groups only have a single |
| 522 | // node as the choice for the second member |
| 523 | // address that one first. |
| 524 | if hasSingleSecondNode(lgrps[k]) == true { |
| 525 | ssn = true |
| 526 | // Add this pairing to the groups |
| 527 | grps = addNode(grps, lgrps[k][0], lgrps[k][1]) |
| 528 | // Since this node is now used, remove it from all |
| 529 | // remaining tenative groups |
| 530 | lgrps = removeNode(lgrps, lgrps[k][1]) |
| 531 | // Now remove this group completely since |
| 532 | // it's been addressed |
| 533 | lgrps = append(lgrps[:k],lgrps[k+1:]...) |
| 534 | break |
| 535 | } |
| 536 | } |
| 537 | if ssn == false { |
| 538 | break |
| 539 | } |
| 540 | } |
| 541 | // Now adress one of the remaining groups |
| 542 | if len(lgrps) == 0 { |
| 543 | break // Nothing left to do, exit the loop |
| 544 | } |
| 545 | grps = addNode(grps, lgrps[0][0], lgrps[0][1]) |
| 546 | lgrps = removeNode(lgrps, lgrps[0][1]) |
| 547 | lgrps = append(lgrps[:0],lgrps[1:]...) |
| 548 | } |
| 549 | return grps |
| 550 | } |
| 551 | |
| 552 | func groupPods1(pods []*rwPod) [][]*rwPod { |
| 553 | var rtrn [][]*rwPod |
| 554 | var podCt int = len(pods) |
| 555 | |
| 556 | rtrn,pods = groupIntersectingPods1(pods, podCt) |
| 557 | // There are several outcomes here |
| 558 | // 1) All pods have been paired and we're done |
| 559 | // 2) Some un-allocated pods remain |
| 560 | // 2.a) All groups have been started |
| 561 | // 2.b) Not all groups have been started |
| 562 | if len(pods) == 0 { |
| 563 | return rtrn |
| 564 | } else if len(rtrn) == podCt >> 1 { // All groupings started |
| 565 | // Allocate the remaining (presumably empty) pods to the started groups |
| 566 | return groupRemainingPods1(rtrn, pods) |
| 567 | } else { // Some groupings started |
| 568 | // Start empty groups with remaining pods |
| 569 | // each grouping is on a different server then |
| 570 | // allocate remaining pods. |
| 571 | rtrn, pods = startRemainingGroups1(rtrn, pods, podCt) |
| 572 | return groupRemainingPods1(rtrn, pods) |
| 573 | } |
| 574 | } |
| 575 | |
| 576 | func groupPods(pods []*rwPod) [][]*rwPod { |
| 577 | var rtrn [][]*rwPod |
| 578 | var pd []*podTrack |
| 579 | |
| 580 | // Tracking of the grouping process |
| 581 | for k,_ := range pods { |
| 582 | pd = append(pd, &podTrack{pods[k],false}) |
| 583 | } |
| 584 | |
| 585 | |
| 586 | rtrn,pd = groupIntersectingPods(pd) |
| 587 | // There are several outcomes here |
| 588 | // 1) All pods have been paired and we're done |
| 589 | // 2) Some un-allocated pods remain |
| 590 | // 2.a) All groups have been started |
| 591 | // 2.b) Not all groups have been started |
| 592 | if unallocPodCount(pd) == 0 { |
| 593 | return rtrn |
| 594 | } else if len(rtrn) == len(pd) >> 1 { // All groupings started |
| 595 | // Allocate the remaining (presumably empty) pods to the started groups |
| 596 | return groupRemainingPods(rtrn, pd) |
| 597 | } else { // Some groupings started |
| 598 | // Start empty groups with remaining pods |
| 599 | // each grouping is on a different server then |
| 600 | // allocate remaining pods. |
| 601 | rtrn, pd = startRemainingGroups(rtrn, pd) |
| 602 | return groupRemainingPods(rtrn, pd) |
| 603 | } |
| 604 | |
| 605 | |
| 606 | // Establish groupings of non-empty pods that have overlapping devices. |
| 607 | for k,_ := range pd { |
| 608 | if pd[k].dn == true { // Already processed? |
| 609 | //log.Debugf("%s already processed", pd[k].pod.name) |
| 610 | continue |
| 611 | } |
| 612 | if len(pd[k].pod.devIds) == 0 { // Ignore pods with no devices |
| 613 | ////log.Debugf("%s empty pod", pd[k].pod.name) |
| 614 | continue |
| 615 | } |
| 616 | // Start a pod group with this pod |
| 617 | var grp []*rwPod |
| 618 | grp = append(grp, pd[k].pod) |
| 619 | pd[k].dn = true |
| 620 | //log.Debugf("Creating new group %s", pd[k].pod.name) |
| 621 | // Find the peer pod based on device overlap |
| 622 | // It's ok if one isn't found, an empty one will be used instead |
| 623 | for k1,_ := range pd { |
| 624 | if pd[k1].dn == true { // Skip over eliminated pods |
| 625 | //log.Debugf("%s eliminated pod", pd[k1].pod.name) |
| 626 | continue |
| 627 | } |
| 628 | if len(pd[k1].pod.devIds) == 0 { // Skip pods with no devices |
| 629 | //log.Debugf("%s empty pod", pd[k1].pod.name) |
| 630 | continue |
| 631 | } |
| 632 | if intersect(pd[k].pod.devIds, pd[k1].pod.devIds) == true { |
| 633 | //log.Debugf("intersection found %s:%s", pd[k].pod.name, pd[k1].pod.name) |
| 634 | pd[k1].dn = true |
| 635 | grp = append(grp, pd[k1].pod) |
| 636 | break |
| 637 | } |
| 638 | } |
| 639 | rtrn = append(rtrn, grp) |
| 640 | //log.Debugf("Added group %s", grp[0].name) |
| 641 | } |
| 642 | // Now find any grouping without 2 members and assign one of the |
| 643 | // pods with no devices and on a different server to it. |
| 644 | // If there are no pods with no devices left before all |
| 645 | // groups are filled report an exception but leave one of the |
| 646 | // groups with only one pod. |
| 647 | for k,_ := range rtrn { |
| 648 | if len(rtrn[k]) < 2 { |
| 649 | for k2,_ := range pd { |
| 650 | if pd[k2].dn == true { |
| 651 | continue |
| 652 | } |
| 653 | // There should be only empty pods here |
| 654 | if len(pd[k2].pod.devIds) != 0 { |
| 655 | log.Error("Non empty pod found where empty pod was expected") |
| 656 | continue |
| 657 | } |
| 658 | if pd[k2].pod.node == rtrn[k][0].node { |
| 659 | //log.Error("Pods aren't on different servers, continuing") |
| 660 | continue |
| 661 | } |
| 662 | // Add this empty and unused pod to the group |
| 663 | //log.Debugf("Adding empty pod %s", pd[k2].pod.name) |
| 664 | rtrn[k] = append(rtrn[k], pd[k2].pod) |
| 665 | pd[k2].dn = true |
| 666 | break |
| 667 | } |
| 668 | } |
| 669 | } |
| 670 | return rtrn |
| 671 | } |
| 672 | |
| 673 | func intersect(d1 map[string]struct{}, d2 map[string]struct{}) bool { |
| 674 | for k,_ := range d1 { |
| 675 | if _,ok := d2[k]; ok == true { |
| 676 | return true |
| 677 | } |
| 678 | } |
| 679 | return false |
| 680 | } |
| 681 | |
| 682 | func setConnection(client pb.ConfigurationClient, backend string, connection string, addr string, port uint64) { |
| 683 | log.Debugf("Configuring backend %s : connection %s\n\n", backend, connection) |
| 684 | cnf := &pb.Conn{Server:"grpc_command",Cluster:"vcore",Backend:backend, |
| 685 | Connection:connection,Addr:addr, |
| 686 | Port:port} |
| 687 | if res, err := client.SetConnection(context.Background(), cnf); err != nil { |
| 688 | log.Debugf("failed SetConnection RPC call: %s", err) |
| 689 | } else { |
| 690 | log.Debugf("Result: %v", res) |
| 691 | } |
| 692 | } |
| 693 | |
| 694 | func setAffinity(client pb.ConfigurationClient, ids map[string]struct{}, backend string) { |
| 695 | log.Debugf("Configuring backend %s : affinities \n", backend) |
| 696 | aff := &pb.Affinity{Router:"vcore",Route:"dev_manager",Cluster:"vcore",Backend:backend} |
| 697 | for k,_ := range ids { |
| 698 | log.Debugf("Setting affinity for id %s", k) |
| 699 | aff.Id = k |
| 700 | if res, err := client.SetAffinity(context.Background(), aff); err != nil { |
| 701 | log.Debugf("failed affinity RPC call: %s", err) |
| 702 | } else { |
| 703 | log.Debugf("Result: %v", res) |
| 704 | } |
| 705 | } |
| 706 | } |
| 707 | |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 708 | func getBackendForCore(coreId string, coreGroups [][]*rwPod) string { |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 709 | for _,v := range coreGroups { |
| 710 | for _,v2 := range v { |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 711 | if v2.name == coreId { |
| 712 | return v2.backend |
| 713 | } |
| 714 | } |
| 715 | } |
| 716 | log.Errorf("No backend found for core %s\n", coreId) |
| 717 | return "" |
| 718 | } |
| 719 | |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 720 | func monitorDiscovery(client pb.ConfigurationClient, |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 721 | ch <-chan *ic.InterContainerMessage, |
| 722 | coreGroups [][]*rwPod) { |
| 723 | var id map[string]struct{} = make(map[string]struct{}) |
| 724 | |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 725 | select { |
| 726 | case msg := <-ch: |
| 727 | log.Debugf("Received a device discovery notification") |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 728 | device := &ic.DeviceDiscovered{} |
| 729 | if err := ptypes.UnmarshalAny(msg.Body, device); err != nil { |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 730 | log.Errorf("Could not unmarshal received notification %v", msg) |
| 731 | } else { |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 732 | // Set the affinity of the discovered device. |
| 733 | if be := getBackendForCore(device.Id, coreGroups); be != "" { |
| 734 | id[device.Id]=struct{}{} |
| 735 | setAffinity(client, id, be) |
| 736 | } else { |
| 737 | log.Error("Cant use an empty string as a backend name") |
| 738 | } |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 739 | } |
| 740 | break |
| 741 | } |
| 742 | } |
| 743 | |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 744 | func startDiscoveryMonitor(client pb.ConfigurationClient, |
| 745 | coreGroups [][]*rwPod) error { |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 746 | var ch <-chan *ic.InterContainerMessage |
| 747 | // Connect to kafka for discovery events |
| 748 | topic := &kafka.Topic{Name: "AffinityRouter"} |
| 749 | kc,err := newKafkaClient("sarama", "kafka", 9092, "arouterd") |
| 750 | kc.Start() |
| 751 | |
| 752 | if ch, err = kc.Subscribe(topic); err != nil { |
| 753 | log.Error("Could not subscribe to the 'AffinityRouter' channel, discovery disabled") |
| 754 | return err |
| 755 | } |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 756 | go monitorDiscovery(client, ch, coreGroups) |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 757 | return nil |
| 758 | } |
| 759 | |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 760 | // Determines which items in core groups |
| 761 | // have changed based on the list provided |
| 762 | // and returns a coreGroup with only the changed |
| 763 | // items and a pod list with the new items |
| 764 | func getAddrDiffs(coreGroups [][]*rwPod, rwPods []*rwPod) ([][]*rwPod, []*rwPod) { |
| 765 | var nList []*rwPod |
| 766 | var rtrn [][]*rwPod = make([][]*rwPod, nPods>>1) |
| 767 | var ipAddrs map[string]struct{} = make(map[string]struct{}) |
| 768 | |
| 769 | log.Debug("Get addr diffs") |
| 770 | |
| 771 | // Start with an empty array |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 772 | for k,_ := range rtrn { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 773 | rtrn[k] = make([]*rwPod, 2) |
| 774 | } |
| 775 | |
| 776 | // Build a list with only the new items |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 777 | for _,v := range rwPods { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 778 | if hasIpAddr(coreGroups, v.ipAddr) == false { |
| 779 | nList = append(nList, v) |
| 780 | } |
| 781 | ipAddrs[v.ipAddr] = struct{}{} // for the search below |
| 782 | } |
| 783 | |
| 784 | // Now build the coreGroups with only the changed items |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 785 | for k1,v1 := range coreGroups { |
| 786 | for k2,v2 := range v1 { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 787 | if _,ok := ipAddrs[v2.ipAddr]; ok == false { |
| 788 | rtrn[k1][k2] = v2 |
| 789 | } |
| 790 | } |
| 791 | } |
| 792 | return rtrn, nList |
| 793 | } |
| 794 | |
| 795 | // Figure out where best to put the new pods |
| 796 | // in the coreGroup array based on the old |
| 797 | // pods being replaced. The criteria is that |
| 798 | // the new pod be on the same server as the |
| 799 | // old pod was. |
| 800 | func reconcileAddrDiffs(coreGroupDiffs [][]*rwPod, rwPodDiffs []*rwPod) ([][]*rwPod) { |
| 801 | var srvrs map[string][]*rwPod = make(map[string][]*rwPod) |
| 802 | |
| 803 | log.Debug("Reconciling diffs") |
| 804 | log.Debug("Building server list") |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 805 | for _,v := range rwPodDiffs { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 806 | log.Debugf("Adding %v to the server list", *v) |
| 807 | srvrs[v.node] = append(srvrs[v.node], v) |
| 808 | } |
| 809 | |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 810 | for k1,v1 := range coreGroupDiffs { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 811 | log.Debugf("k1:%v, v1:%v", k1,v1) |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 812 | for k2,v2 := range v1 { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 813 | log.Debugf("k2:%v, v2:%v", k2,v2) |
| 814 | if v2 == nil { // Nothing to do here |
| 815 | continue |
| 816 | } |
| 817 | if _,ok := srvrs[v2.node]; ok == true { |
| 818 | coreGroupDiffs[k1][k2] = srvrs[v2.node][0] |
| 819 | if len(srvrs[v2.node]) > 1 { // remove one entry from the list |
| 820 | srvrs[v2.node] = append(srvrs[v2.node][:0], srvrs[v2.node][1:]...) |
| 821 | } else { // Delete the endtry from the map |
| 822 | delete(srvrs, v2.node) |
| 823 | } |
| 824 | } else { |
| 825 | log.Error("This should never happen, node appears to have changed names") |
| 826 | // attempt to limp along by keeping this old entry |
| 827 | } |
| 828 | } |
| 829 | } |
| 830 | |
| 831 | return coreGroupDiffs |
| 832 | } |
| 833 | |
| 834 | func applyAddrDiffs(client pb.ConfigurationClient, coreGroups [][]*rwPod, rwPods []*rwPod) { |
| 835 | var newEntries [][]*rwPod |
| 836 | |
| 837 | log.Debug("Applying diffs") |
| 838 | newEntries = reconcileAddrDiffs(getAddrDiffs(coreGroups, rwPods)) |
| 839 | |
| 840 | // Now replace the information in coreGropus with the new |
| 841 | // entries and then reconcile the device ids on the core |
| 842 | // that's in the new entry with the device ids of it's |
| 843 | // active-active peer. |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 844 | for k1,v1 := range coreGroups { |
| 845 | for k2,v2 := range v1 { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 846 | if newEntries[k1][k2] != nil { |
| 847 | // TODO: Missing is the case where bothe the primary |
| 848 | // and the secondary core crash and come back. |
| 849 | // Pull the device ids from the active-active peer |
| 850 | ids := queryPodDeviceIds(coreGroups[k1][k2^1]) |
| 851 | if len(ids) != 0 { |
| 852 | if reconcilePodDeviceIds(newEntries[k1][k2], ids) == false { |
| 853 | log.Errorf("Attempt to reconcile ids on pod %v failed",newEntries[k1][k2]) |
| 854 | } |
| 855 | } |
| 856 | // Send the affininty router new connection information |
| 857 | setConnection(client, v2.backend, v2.connection, newEntries[k1][k2].ipAddr, 50057) |
| 858 | // Copy the new entry information over |
| 859 | coreGroups[k1][k2].ipAddr = newEntries[k1][k2].ipAddr |
| 860 | coreGroups[k1][k2].name = newEntries[k1][k2].name |
| 861 | coreGroups[k1][k2].devIds = ids |
| 862 | } |
| 863 | } |
| 864 | } |
| 865 | } |
| 866 | |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 867 | func startCoreMonitor(client pb.ConfigurationClient, |
| 868 | clientset *kubernetes.Clientset, |
| 869 | coreFltr *regexp.Regexp, |
| 870 | coreGroups [][]*rwPod) error { |
| 871 | // Now that initial allocation has been completed, monitor the pods |
| 872 | // for IP changes |
| 873 | // The main loop needs to do the following: |
| 874 | // 1) Periodically query the pods and filter out |
| 875 | // the vcore ones |
| 876 | // 2) Validate that the pods running are the same |
| 877 | // as the previous check |
| 878 | // 3) Validate that the IP addresses are the same |
| 879 | // as the last check. |
| 880 | // If the pod name(s) ha(s/ve) changed then remove |
| 881 | // the unused pod names and add in the new pod names |
| 882 | // maintaining the cluster/backend information. |
| 883 | // If an IP address has changed (which shouldn't |
| 884 | // happen unless a pod is re-started) it should get |
| 885 | // caught by the pod name change. |
| 886 | for { |
| 887 | time.Sleep(10 * time.Second) // Wait a while |
| 888 | // Get the rw core list from k8s |
| 889 | rwPods := getRwPods(clientset, coreFltr) |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 890 | queryDeviceIds(rwPods) |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 891 | // If we didn't get 2n+1 pods then wait since |
| 892 | // something is down and will hopefully come |
| 893 | // back up at some point. |
| 894 | // TODO: remove the 6 pod hardcoding |
| 895 | if len(rwPods) != 6 { |
| 896 | continue |
| 897 | } |
| 898 | // We have all pods, check if any IP addresses |
| 899 | // have changed. |
| 900 | for _,v := range rwPods { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 901 | if hasIpAddr(coreGroups, v.ipAddr) == false { |
| 902 | log.Debug("Address has changed...") |
| 903 | applyAddrDiffs(client, coreGroups, rwPods) |
| 904 | |
| 905 | } |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 906 | } |
| 907 | } |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 908 | } |
| 909 | |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 910 | func hasIpAddr(coreGroups [][]*rwPod, ipAddr string) bool { |
sslobodr | 6c1689c | 2019-01-24 07:31:15 -0500 | [diff] [blame] | 911 | for _,v1 := range coreGroups { |
| 912 | for _,v2 := range v1 { |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 913 | if v2.ipAddr == ipAddr { |
| 914 | return true |
| 915 | } |
| 916 | } |
| 917 | } |
| 918 | return false |
| 919 | } |
| 920 | |
| 921 | |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 922 | func main() { |
| 923 | // This is currently hard coded to a cluster with 3 servers |
| 924 | //var connections map[string]configConn = make(map[string]configConn) |
| 925 | //var rwCorePodsPrev map[string]rwPod = make(map[string]rwPod) |
| 926 | var rwCoreNodesPrev map[string][]rwPod = make(map[string][]rwPod) |
| 927 | var firstTime bool = true |
| 928 | var err error |
| 929 | var conn *grpc.ClientConn |
| 930 | |
| 931 | |
| 932 | // Set up the regular expression to identify the voltha cores |
| 933 | coreFltr := regexp.MustCompile(`rw-core[0-9]-`) |
| 934 | |
| 935 | // Set up logging |
| 936 | if _, err := log.SetDefaultLogger(log.JSON, 0, nil); err != nil { |
| 937 | log.With(log.Fields{"error": err}).Fatal("Cannot setup logging") |
| 938 | } |
| 939 | |
| 940 | // Set up kubernetes api |
| 941 | clientset := k8sClientSet() |
| 942 | |
| 943 | // Connect to the affinity router and set up the client |
| 944 | conn, err = connect("localhost:55554") // This is a sidecar container so communicating over localhost |
| 945 | if err != nil { |
| 946 | panic(err.Error()) |
| 947 | } |
| 948 | client := pb.NewConfigurationClient(conn) |
| 949 | |
| 950 | // Get the voltha rw-core podes |
| 951 | rwPods := getRwPods(clientset, coreFltr) |
| 952 | |
| 953 | // Fetch the devices held by each running core |
sslobodr | e7ce71d | 2019-01-22 16:21:45 -0500 | [diff] [blame] | 954 | queryDeviceIds(rwPods) |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 955 | |
| 956 | // For debugging... comment out l8r |
| 957 | for _,v := range rwPods { |
| 958 | log.Debugf("Pod list %v", *v) |
| 959 | } |
| 960 | |
| 961 | coreGroups := groupPods1(rwPods) |
| 962 | |
| 963 | |
| 964 | // Assign the groupings to the the backends and connections |
| 965 | for k,_ := range coreGroups { |
| 966 | for k1,_ := range coreGroups[k] { |
| 967 | coreGroups[k][k1].backend = "vcore"+strconv.Itoa(k+1) |
| 968 | coreGroups[k][k1].connection = "vcore"+strconv.Itoa(k+1)+strconv.Itoa(k1+1) |
| 969 | } |
| 970 | } |
| 971 | log.Debug("Core gouping completed") |
| 972 | |
| 973 | // TODO: Debugging code, comment out for production |
| 974 | for k,v := range coreGroups { |
| 975 | for k2,v2 := range v { |
| 976 | log.Debugf("Core group %d,%d: %v", k, k2, v2) |
| 977 | } |
| 978 | } |
| 979 | log.Debug("Setting affinities") |
| 980 | // Now set the affinities for exising devices in the cores |
| 981 | for _,v := range coreGroups { |
| 982 | setAffinity(client, v[0].devIds, v[0].backend) |
| 983 | setAffinity(client, v[1].devIds, v[1].backend) |
| 984 | } |
| 985 | log.Debug("Setting connections") |
| 986 | // Configure the backeds based on the calculated core groups |
| 987 | for _,v := range coreGroups { |
| 988 | setConnection(client, v[0].backend, v[0].connection, v[0].ipAddr, 50057) |
| 989 | setConnection(client, v[1].backend, v[1].connection, v[1].ipAddr, 50057) |
| 990 | } |
| 991 | |
| 992 | log.Debug("Starting discovery monitoring") |
sslobodr | 38afd0d | 2019-01-21 12:31:46 -0500 | [diff] [blame] | 993 | startDiscoveryMonitor(client, coreGroups) |
sslobodr | 16e41bc | 2019-01-18 16:22:21 -0500 | [diff] [blame] | 994 | |
| 995 | log.Debugf("Starting core monitoring") |
| 996 | startCoreMonitor(client, clientset, coreFltr, coreGroups) // Never returns |
| 997 | return |
| 998 | |
| 999 | |
| 1000 | // The main loop needs to do the following: |
| 1001 | // 1) Periodically query the pods and filter out |
| 1002 | // the vcore ones |
| 1003 | // 2) Validate that the pods running are the same |
| 1004 | // as the previous check |
| 1005 | // 3) Validate that the IP addresses are the same |
| 1006 | // as the last check. |
| 1007 | // If the pod name(s) ha(s/ve) changed then remove |
| 1008 | // the unused pod names and add in the new pod names |
| 1009 | // maintaining the cluster/backend information. |
| 1010 | // If an IP address has changed (which shouldn't |
| 1011 | // happen unless a pod is re-started) it should get |
| 1012 | // caught by the pod name change. |
| 1013 | for { |
| 1014 | var rwCorePods map[string]rwPod = make(map[string]rwPod) |
| 1015 | var rwCoreNodes map[string][]rwPod = make(map[string][]rwPod) |
| 1016 | pods, err := clientset.CoreV1().Pods("").List(metav1.ListOptions{}) |
| 1017 | if err != nil { |
| 1018 | panic(err.Error()) |
| 1019 | } |
| 1020 | log.Debugf("There are %d pods in the cluster\n", len(pods.Items)) |
| 1021 | |
| 1022 | /* |
| 1023 | for k,v := range pods.Items { |
| 1024 | if v.Namespace == "voltha" && coreFltr.MatchString(v.Name) { |
| 1025 | fmt.Printf("Namespace: %s, PodName: %s, PodIP: %s, Host: %s\n", v.Namespace, v.Name, |
| 1026 | v.Status.PodIP, v.Spec.NodeName) |
| 1027 | //fmt.Printf("Pod %v,%v\n\n\n",k,v) |
| 1028 | _ = k |
| 1029 | // Add this pod to the core structure. |
| 1030 | if firstTime == true { |
| 1031 | rwCorePodsPrev[v.Name] = rwPod{name:v.Name,node:v.Spec.NodeName} |
| 1032 | rwCoreNodesPrev[v.Spec.NodeName] = |
| 1033 | append(rwCoreNodesPrev[v.Spec.NodeName], rwPod{name:v.Name,node:v.Spec.NodeName}) |
| 1034 | } |
| 1035 | rwCorePods[v.Name] = rwPod{v.Name,v.Status.PodIP,v.Spec.NodeName, "", ""} |
| 1036 | rwCoreNodes[v.Spec.NodeName] = |
| 1037 | append(rwCoreNodes[v.Spec.NodeName], rwPod{v.Name,v.Status.PodIP,v.Spec.NodeName,"",""}) |
| 1038 | } |
| 1039 | } |
| 1040 | */ |
| 1041 | |
| 1042 | if len(rwCorePods) != 6 { |
| 1043 | continue |
| 1044 | } |
| 1045 | |
| 1046 | //fmt.Printf("Pod map: %v\n", rwCorePods) |
| 1047 | //fmt.Printf("Pod map2: %v\n", rwCoreNodes) |
| 1048 | |
| 1049 | // Examples for error handling: |
| 1050 | // - Use helper functions like e.g. errors.IsNotFound() |
| 1051 | // - And/or cast to StatusError and use its properties like e.g. ErrStatus.Message |
| 1052 | /* |
| 1053 | _, err = clientset.CoreV1().Pods("default").Get("example-xxxxx", metav1.GetOptions{}) |
| 1054 | if errors.IsNotFound(err) { |
| 1055 | fmt.Printf("Pod not found\n") |
| 1056 | } else if statusError, isStatus := err.(*errors.StatusError); isStatus { |
| 1057 | fmt.Printf("Error getting pod %v\n", statusError.ErrStatus.Message) |
| 1058 | } else if err != nil { |
| 1059 | panic(err.Error()) |
| 1060 | } else { |
| 1061 | fmt.Printf("Found pod\n") |
| 1062 | } |
| 1063 | */ |
| 1064 | // Set the association to backends and connections only once. |
| 1065 | // TODO: This needs to be reworked for when a pod crashes |
| 1066 | // and it's name changes. |
| 1067 | if firstTime == true { |
| 1068 | be := 1 |
| 1069 | for k,_ := range rwCoreNodesPrev { // Each node has 2 cores running on it |
| 1070 | // Use a pretty dumb distribution algorithm. |
| 1071 | log.Debugf("Processing core node %s:%d\n", k,be) |
| 1072 | rwCoreNodesPrev[k][0].backend = "vcore"+strconv.Itoa(be) |
| 1073 | rwCoreNodesPrev[k][0].connection = "vcore"+strconv.Itoa(be)+strconv.Itoa(1) |
| 1074 | rwCoreNodesPrev[k][1].backend = "vcore"+strconv.Itoa(be%3+1) |
| 1075 | rwCoreNodesPrev[k][1].connection = "vcore"+strconv.Itoa(be%3+1)+strconv.Itoa(2) |
| 1076 | be++ |
| 1077 | } |
| 1078 | } |
| 1079 | |
| 1080 | log.Debugf("Backend Allocation: %v",rwCoreNodesPrev) |
| 1081 | // Compare the current node IPs with the previous node IPs and if they differ |
| 1082 | // then set the new one and send the command to configure the router with the |
| 1083 | // new backend connection. |
| 1084 | for k,v := range rwCoreNodesPrev { |
| 1085 | if rwCoreNodes[k][0].ipAddr != rwCoreNodesPrev[k][0].ipAddr { |
| 1086 | log.Debugf("Configuring backend %s : connection %s\n\n", v[0].backend, v[0].connection) |
| 1087 | cnf := &pb.Conn{Server:"grpc_command",Cluster:"vcore",Backend:rwCoreNodesPrev[k][0].backend, |
| 1088 | Connection:rwCoreNodesPrev[k][0].connection,Addr:rwCoreNodes[k][0].ipAddr, |
| 1089 | Port:50057} |
| 1090 | if res, err := client.SetConnection(context.Background(), cnf); err != nil { |
| 1091 | log.Debugf("failed SetConnection RPC call: %s", err) |
| 1092 | } else { |
| 1093 | log.Debugf("Result: %v", res) |
| 1094 | rwCoreNodesPrev[k][0].ipAddr = rwCoreNodes[k][0].ipAddr |
| 1095 | } |
| 1096 | } |
| 1097 | if rwCoreNodes[k][1].ipAddr != rwCoreNodesPrev[k][1].ipAddr { |
| 1098 | log.Debugf("Configuring backend %s : connection %s\n\n", v[1].backend, v[1].connection) |
| 1099 | cnf := &pb.Conn{Server:"grpc_command",Cluster:"vcore",Backend:rwCoreNodesPrev[k][1].backend, |
| 1100 | Connection:rwCoreNodesPrev[k][1].connection,Addr:rwCoreNodes[k][1].ipAddr, |
| 1101 | Port:50057} |
| 1102 | if res, err := client.SetConnection(context.Background(), cnf); err != nil { |
| 1103 | log.Debugf("failed SetConnection RPC call: %s", err) |
| 1104 | } else { |
| 1105 | log.Debugf("Result: %v", res) |
| 1106 | rwCoreNodesPrev[k][1].ipAddr = rwCoreNodes[k][1].ipAddr |
| 1107 | } |
| 1108 | } |
| 1109 | } |
| 1110 | |
| 1111 | |
| 1112 | fmt.Printf("The structure for setting the connections is: %v\n", rwCoreNodesPrev) |
| 1113 | firstTime = false |
| 1114 | |
| 1115 | // Now make the API calls |
| 1116 | time.Sleep(10 * time.Second) |
| 1117 | } |
| 1118 | conn.Close() |
| 1119 | |
| 1120 | } |
| 1121 | |