blob: 0206ee042b31e8c2be912378d2d0225ddeedf896 [file] [log] [blame]
khenaidoob9203542018-09-17 22:56:37 -04001/*
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 */
npujar1d86a522019-11-14 17:11:16 +053016
khenaidoob9203542018-09-17 22:56:37 -040017package core
18
19import (
20 "context"
Thomas Lee Se5a44012019-11-07 20:32:24 +053021 "fmt"
Scott Baker2d87ee32020-03-03 13:04:01 -080022 "sync"
npujar1d86a522019-11-14 17:11:16 +053023 "time"
24
sbarbari17d7e222019-11-05 10:02:29 -050025 "github.com/opencord/voltha-go/db/model"
khenaidoob9203542018-09-17 22:56:37 -040026 "github.com/opencord/voltha-go/rw_core/config"
serkant.uluderya2ae470f2020-01-21 11:13:09 -080027 "github.com/opencord/voltha-lib-go/v3/pkg/db"
28 "github.com/opencord/voltha-lib-go/v3/pkg/db/kvstore"
29 grpcserver "github.com/opencord/voltha-lib-go/v3/pkg/grpc"
30 "github.com/opencord/voltha-lib-go/v3/pkg/kafka"
31 "github.com/opencord/voltha-lib-go/v3/pkg/log"
32 "github.com/opencord/voltha-lib-go/v3/pkg/probe"
33 "github.com/opencord/voltha-protos/v3/go/voltha"
khenaidoob9203542018-09-17 22:56:37 -040034 "google.golang.org/grpc"
khenaidoob3244212019-08-27 14:32:27 -040035 "google.golang.org/grpc/codes"
36 "google.golang.org/grpc/status"
khenaidoob9203542018-09-17 22:56:37 -040037)
38
npujar1d86a522019-11-14 17:11:16 +053039// Core represent read,write core attributes
khenaidoob9203542018-09-17 22:56:37 -040040type Core struct {
npujar1d86a522019-11-14 17:11:16 +053041 instanceID string
khenaidoob9203542018-09-17 22:56:37 -040042 deviceMgr *DeviceManager
43 logicalDeviceMgr *LogicalDeviceManager
44 grpcServer *grpcserver.GrpcServer
Richard Jankowskidbab94a2018-12-06 16:20:25 -050045 grpcNBIAPIHandler *APIHandler
khenaidoo2c6a0992019-04-29 13:46:56 -040046 adapterMgr *AdapterManager
khenaidoob9203542018-09-17 22:56:37 -040047 config *config.RWCoreFlags
npujar467fe752020-01-16 20:17:45 +053048 kmp kafka.InterContainerProxy
khenaidoo92e62c52018-10-03 14:02:54 -040049 clusterDataRoot model.Root
50 localDataRoot model.Root
khenaidoob9203542018-09-17 22:56:37 -040051 clusterDataProxy *model.Proxy
52 localDataProxy *model.Proxy
Scott Baker2d87ee32020-03-03 13:04:01 -080053 exitChannel chan struct{}
54 stopOnce sync.Once
Richard Jankowskie4d77662018-10-17 13:53:21 -040055 kvClient kvstore.Client
Girish Kumar4d3887d2019-11-22 14:22:05 +000056 backend db.Backend
khenaidoo43c82122018-11-22 18:38:28 -050057 kafkaClient kafka.Client
khenaidoo2c6a0992019-04-29 13:46:56 -040058 deviceOwnership *DeviceOwnership
khenaidoob9203542018-09-17 22:56:37 -040059}
60
61func init() {
npujar1d86a522019-11-14 17:11:16 +053062 _, err := log.AddPackage(log.JSON, log.WarnLevel, nil)
63 if err != nil {
64 log.Errorw("unable-to-register-package-to-the-log-map", log.Fields{"error": err})
65 }
khenaidoob9203542018-09-17 22:56:37 -040066}
67
npujar1d86a522019-11-14 17:11:16 +053068// NewCore creates instance of rw core
Thomas Lee Se5a44012019-11-07 20:32:24 +053069func NewCore(ctx context.Context, id string, cf *config.RWCoreFlags, kvClient kvstore.Client, kafkaClient kafka.Client) *Core {
khenaidoob9203542018-09-17 22:56:37 -040070 var core Core
npujar1d86a522019-11-14 17:11:16 +053071 core.instanceID = id
Scott Baker2d87ee32020-03-03 13:04:01 -080072 core.exitChannel = make(chan struct{})
khenaidoob9203542018-09-17 22:56:37 -040073 core.config = cf
Richard Jankowskie4d77662018-10-17 13:53:21 -040074 core.kvClient = kvClient
khenaidoo43c82122018-11-22 18:38:28 -050075 core.kafkaClient = kafkaClient
Richard Jankowskie4d77662018-10-17 13:53:21 -040076
Girish Kumar4d3887d2019-11-22 14:22:05 +000077 // Configure backend to push Liveness Status at least every (cf.LiveProbeInterval / 2) seconds
78 // so as to avoid trigger of Liveness check (due to Liveness timeout) when backend is alive
79 livenessChannelInterval := cf.LiveProbeInterval / 2
80
Richard Jankowskie4d77662018-10-17 13:53:21 -040081 // Setup the KV store
Girish Kumar4d3887d2019-11-22 14:22:05 +000082 core.backend = db.Backend{
83 Client: kvClient,
84 StoreType: cf.KVStoreType,
85 Host: cf.KVStoreHost,
86 Port: cf.KVStorePort,
87 Timeout: cf.KVStoreTimeout,
88 LivenessChannelInterval: livenessChannelInterval,
89 PathPrefix: cf.KVStoreDataPrefix}
90 core.clusterDataRoot = model.NewRoot(&voltha.Voltha{}, &core.backend)
91 core.localDataRoot = model.NewRoot(&voltha.CoreInstance{}, &core.backend)
khenaidoob9203542018-09-17 22:56:37 -040092 return &core
93}
94
npujar1d86a522019-11-14 17:11:16 +053095// Start brings up core services
Thomas Lee Se5a44012019-11-07 20:32:24 +053096func (core *Core) Start(ctx context.Context) error {
David K. Bainbridgeb4a9ab02019-09-20 15:12:16 -070097
98 // If the context has a probe then fetch it and register our services
99 var p *probe.Probe
100 if value := ctx.Value(probe.ProbeContextKey); value != nil {
101 if _, ok := value.(*probe.Probe); ok {
102 p = value.(*probe.Probe)
103 p.RegisterService(
104 "message-bus",
105 "kv-store",
106 "device-manager",
107 "logical-device-manager",
108 "adapter-manager",
109 "grpc-service",
110 )
111 }
112 }
113
npujar1d86a522019-11-14 17:11:16 +0530114 log.Info("starting-core-services", log.Fields{"coreId": core.instanceID})
khenaidoob3244212019-08-27 14:32:27 -0400115
116 // Wait until connection to KV Store is up
117 if err := core.waitUntilKVStoreReachableOrMaxTries(ctx, core.config.MaxConnectionRetries, core.config.ConnectionRetryInterval); err != nil {
118 log.Fatal("Unable-to-connect-to-KV-store")
119 }
David K. Bainbridgeb4a9ab02019-09-20 15:12:16 -0700120 if p != nil {
121 p.UpdateStatus("kv-store", probe.ServiceStatusRunning)
122 }
Thomas Lee Se5a44012019-11-07 20:32:24 +0530123 var err error
124
npujar467fe752020-01-16 20:17:45 +0530125 core.clusterDataProxy, err = core.clusterDataRoot.CreateProxy(ctx, "/", false)
Thomas Lee Se5a44012019-11-07 20:32:24 +0530126 if err != nil {
127 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusNotReady)
128 return fmt.Errorf("Failed to create cluster data proxy")
129 }
npujar467fe752020-01-16 20:17:45 +0530130 core.localDataProxy, err = core.localDataRoot.CreateProxy(ctx, "/", false)
Thomas Lee Se5a44012019-11-07 20:32:24 +0530131 if err != nil {
132 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusNotReady)
133 return fmt.Errorf("Failed to create local data proxy")
134 }
khenaidoob3244212019-08-27 14:32:27 -0400135
Scott Bakeree6a0872019-10-29 15:59:52 -0700136 // core.kmp must be created before deviceMgr and adapterMgr, as they will make
137 // private copies of the poiner to core.kmp.
npujar467fe752020-01-16 20:17:45 +0530138 core.initKafkaManager(ctx)
khenaidoob3244212019-08-27 14:32:27 -0400139
khenaidoo631fe542019-05-31 15:44:43 -0400140 log.Debugw("values", log.Fields{"kmp": core.kmp})
Richard Jankowski199fd862019-03-18 14:49:51 -0400141 core.deviceMgr = newDeviceManager(core)
Kent Hagerman16ce36a2019-12-17 13:40:53 -0500142 core.adapterMgr = newAdapterManager(core.clusterDataProxy, core.instanceID, core.kafkaClient, core.deviceMgr)
khenaidooba6b6c42019-08-02 09:11:56 -0400143 core.deviceMgr.adapterMgr = core.adapterMgr
khenaidoo2c6a0992019-04-29 13:46:56 -0400144 core.logicalDeviceMgr = newLogicalDeviceManager(core, core.deviceMgr, core.kmp, core.clusterDataProxy, core.config.DefaultCoreTimeout)
khenaidoo54e0ddf2019-02-27 16:21:33 -0500145
Scott Bakeree6a0872019-10-29 15:59:52 -0700146 // Start the KafkaManager. This must be done after the deviceMgr, adapterMgr, and
147 // logicalDeviceMgr have been created, as once the kmp is started, it will register
148 // the above with the kmp.
149
150 go core.startKafkaManager(ctx,
151 core.config.ConnectionRetryInterval,
152 core.config.LiveProbeInterval,
153 core.config.NotLiveProbeInterval)
khenaidoob3244212019-08-27 14:32:27 -0400154
khenaidoob9203542018-09-17 22:56:37 -0400155 go core.startDeviceManager(ctx)
156 go core.startLogicalDeviceManager(ctx)
157 go core.startGRPCService(ctx)
khenaidoo21d51152019-02-01 13:48:37 -0500158 go core.startAdapterManager(ctx)
Girish Kumar4d3887d2019-11-22 14:22:05 +0000159 go core.monitorKvstoreLiveness(ctx)
khenaidoob9203542018-09-17 22:56:37 -0400160
khenaidoo1ce37ad2019-03-24 22:07:24 -0400161 // Setup device ownership context
npujar1d86a522019-11-14 17:11:16 +0530162 core.deviceOwnership = NewDeviceOwnership(core.instanceID, core.kvClient, core.deviceMgr, core.logicalDeviceMgr,
khenaidoo1ce37ad2019-03-24 22:07:24 -0400163 "service/voltha/owns_device", 10)
164
khenaidoob3244212019-08-27 14:32:27 -0400165 log.Info("core-services-started")
Thomas Lee Se5a44012019-11-07 20:32:24 +0530166 return nil
khenaidoob9203542018-09-17 22:56:37 -0400167}
168
npujar1d86a522019-11-14 17:11:16 +0530169// Stop brings down core services
khenaidoob9203542018-09-17 22:56:37 -0400170func (core *Core) Stop(ctx context.Context) {
Scott Baker2d87ee32020-03-03 13:04:01 -0800171 core.stopOnce.Do(func() {
172 log.Info("stopping-adaptercore")
173 // Signal to the KVStoreMonitor that we are stopping.
174 close(core.exitChannel)
175 // Stop all the started services
176 if core.grpcServer != nil {
177 core.grpcServer.Stop()
178 }
179 if core.logicalDeviceMgr != nil {
180 core.logicalDeviceMgr.stop(ctx)
181 }
182 if core.deviceMgr != nil {
183 core.deviceMgr.stop(ctx)
184 }
185 if core.kmp != nil {
186 core.kmp.Stop()
187 }
188 log.Info("adaptercore-stopped")
189 })
khenaidoob9203542018-09-17 22:56:37 -0400190}
191
khenaidoo631fe542019-05-31 15:44:43 -0400192//startGRPCService creates the grpc service handlers, registers it to the grpc server and starts the server
khenaidoob9203542018-09-17 22:56:37 -0400193func (core *Core) startGRPCService(ctx context.Context) {
194 // create an insecure gserver server
Scott Bakeree6a0872019-10-29 15:59:52 -0700195 core.grpcServer = grpcserver.NewGrpcServer(core.config.GrpcHost, core.config.GrpcPort, nil, false, probe.GetProbeFromContext(ctx))
khenaidoob9203542018-09-17 22:56:37 -0400196 log.Info("grpc-server-created")
197
khenaidoo54e0ddf2019-02-27 16:21:33 -0500198 core.grpcNBIAPIHandler = NewAPIHandler(core)
Richard Jankowski46464e92019-03-05 11:53:55 -0500199 log.Infow("grpc-handler", log.Fields{"core_binding_key": core.config.CoreBindingKey})
Richard Jankowskidbab94a2018-12-06 16:20:25 -0500200 core.logicalDeviceMgr.setGrpcNbiHandler(core.grpcNBIAPIHandler)
khenaidoob9203542018-09-17 22:56:37 -0400201 // Create a function to register the core GRPC service with the GRPC server
202 f := func(gs *grpc.Server) {
203 voltha.RegisterVolthaServiceServer(
204 gs,
Richard Jankowskidbab94a2018-12-06 16:20:25 -0500205 core.grpcNBIAPIHandler,
khenaidoob9203542018-09-17 22:56:37 -0400206 )
207 }
208
209 core.grpcServer.AddService(f)
210 log.Info("grpc-service-added")
211
David K. Bainbridgeb4a9ab02019-09-20 15:12:16 -0700212 /*
213 * Start the GRPC server
214 *
215 * This is a bit sub-optimal here as the grpcServer.Start call does not return (blocks)
216 * until something fails, but we want to send a "start" status update. As written this
217 * means that we are actually sending the "start" status update before the server is
218 * started, which means it is possible that the status is "running" before it actually is.
219 *
220 * This means that there is a small window in which the core could return its status as
221 * ready, when it really isn't.
222 */
223 probe.UpdateStatusFromContext(ctx, "grpc-service", probe.ServiceStatusRunning)
khenaidoob9203542018-09-17 22:56:37 -0400224 log.Info("grpc-server-started")
npujar467fe752020-01-16 20:17:45 +0530225 core.grpcServer.Start(ctx)
David K. Bainbridgeb4a9ab02019-09-20 15:12:16 -0700226 probe.UpdateStatusFromContext(ctx, "grpc-service", probe.ServiceStatusStopped)
khenaidoob9203542018-09-17 22:56:37 -0400227}
228
Scott Bakeree6a0872019-10-29 15:59:52 -0700229// Initialize the kafka manager, but we will start it later
npujar467fe752020-01-16 20:17:45 +0530230func (core *Core) initKafkaManager(ctx context.Context) {
Scott Bakeree6a0872019-10-29 15:59:52 -0700231 log.Infow("initialize-kafka-manager", log.Fields{"host": core.config.KafkaAdapterHost,
khenaidoob9203542018-09-17 22:56:37 -0400232 "port": core.config.KafkaAdapterPort, "topic": core.config.CoreTopic})
Scott Bakeree6a0872019-10-29 15:59:52 -0700233
234 probe.UpdateStatusFromContext(ctx, "message-bus", probe.ServiceStatusPreparing)
235
236 // create the proxy
npujar467fe752020-01-16 20:17:45 +0530237 core.kmp = kafka.NewInterContainerProxy(
khenaidoo43c82122018-11-22 18:38:28 -0500238 kafka.InterContainerHost(core.config.KafkaAdapterHost),
239 kafka.InterContainerPort(core.config.KafkaAdapterPort),
240 kafka.MsgClient(core.kafkaClient),
khenaidoo79232702018-12-04 11:00:41 -0500241 kafka.DefaultTopic(&kafka.Topic{Name: core.config.CoreTopic}),
npujar467fe752020-01-16 20:17:45 +0530242 kafka.DeviceDiscoveryTopic(&kafka.Topic{Name: core.config.AffinityRouterTopic}))
Scott Bakeree6a0872019-10-29 15:59:52 -0700243
244 probe.UpdateStatusFromContext(ctx, "message-bus", probe.ServiceStatusPrepared)
Scott Bakeree6a0872019-10-29 15:59:52 -0700245}
246
247/*
248 * KafkaMonitorThread
249 *
npujar1d86a522019-11-14 17:11:16 +0530250 * Responsible for starting the Kafka Interadapter Proxy and monitoring its liveness
Scott Bakeree6a0872019-10-29 15:59:52 -0700251 * state.
252 *
253 * Any producer that fails to send will cause KafkaInterContainerProxy to
254 * post a false event on its liveness channel. Any producer that succeeds in sending
255 * will cause KafkaInterContainerProxy to post a true event on its liveness
npujar1d86a522019-11-14 17:11:16 +0530256 * channel. Group receivers also update liveness state, and a receiver will typically
Scott Bakeree6a0872019-10-29 15:59:52 -0700257 * indicate a loss of liveness within 3-5 seconds of Kafka going down. Receivers
258 * only indicate restoration of liveness if a message is received. During normal
259 * operation, messages will be routinely produced and received, automatically
260 * indicating liveness state. These routine liveness indications are rate-limited
261 * inside sarama_client.
262 *
263 * This thread monitors the status of KafkaInterContainerProxy's liveness and pushes
264 * that state to the core's readiness probes. If no liveness event has been seen
265 * within a timeout, then the thread will make an attempt to produce a "liveness"
266 * message, which will in turn trigger a liveness event on the liveness channel, true
267 * or false depending on whether the attempt succeeded.
268 *
269 * The gRPC server in turn monitors the state of the readiness probe and will
270 * start issuing UNAVAILABLE response while the probe is not ready.
271 *
272 * startupRetryInterval -- interval between attempts to start
273 * liveProbeInterval -- interval between liveness checks when in a live state
274 * notLiveProbeInterval -- interval between liveness checks when in a notLive state
275 *
276 * liveProbeInterval and notLiveProbeInterval can be configured separately,
277 * though the current default is that both are set to 60 seconds.
278 */
279
Girish Kumar4d3887d2019-11-22 14:22:05 +0000280func (core *Core) startKafkaManager(ctx context.Context, startupRetryInterval time.Duration, liveProbeInterval time.Duration, notLiveProbeInterval time.Duration) {
Scott Bakeree6a0872019-10-29 15:59:52 -0700281 log.Infow("starting-kafka-manager-thread", log.Fields{"host": core.config.KafkaAdapterHost,
282 "port": core.config.KafkaAdapterPort, "topic": core.config.CoreTopic})
283
284 started := false
285 for !started {
286 // If we haven't started yet, then try to start
287 log.Infow("starting-kafka-proxy", log.Fields{})
288 if err := core.kmp.Start(); err != nil {
289 // We failed to start. Delay and then try again later.
290 // Don't worry about liveness, as we can't be live until we've started.
291 probe.UpdateStatusFromContext(ctx, "message-bus", probe.ServiceStatusNotReady)
khenaidoob3244212019-08-27 14:32:27 -0400292 log.Infow("error-starting-kafka-messaging-proxy", log.Fields{"error": err})
Girish Kumar4d3887d2019-11-22 14:22:05 +0000293 time.Sleep(startupRetryInterval)
khenaidoob3244212019-08-27 14:32:27 -0400294 } else {
Scott Bakeree6a0872019-10-29 15:59:52 -0700295 // We started. We only need to do this once.
296 // Next we'll fall through and start checking liveness.
297 log.Infow("started-kafka-proxy", log.Fields{})
298
299 // cannot do this until after the kmp is started
npujar1d86a522019-11-14 17:11:16 +0530300 if err := core.registerAdapterRequestHandlers(ctx, core.instanceID, core.deviceMgr, core.logicalDeviceMgr, core.adapterMgr, core.clusterDataProxy, core.localDataProxy); err != nil {
Scott Bakeree6a0872019-10-29 15:59:52 -0700301 log.Fatal("Failure-registering-adapterRequestHandler")
302 }
303
304 started = true
khenaidoob3244212019-08-27 14:32:27 -0400305 }
khenaidoob9203542018-09-17 22:56:37 -0400306 }
Scott Bakeree6a0872019-10-29 15:59:52 -0700307
308 log.Info("started-kafka-message-proxy")
309
310 livenessChannel := core.kmp.EnableLivenessChannel(true)
311
312 log.Info("enabled-kafka-liveness-channel")
313
Girish Kumar4d3887d2019-11-22 14:22:05 +0000314 timeout := liveProbeInterval
Scott Bakeree6a0872019-10-29 15:59:52 -0700315 for {
316 timeoutTimer := time.NewTimer(timeout)
317 select {
318 case liveness := <-livenessChannel:
319 log.Infow("kafka-manager-thread-liveness-event", log.Fields{"liveness": liveness})
320 // there was a state change in Kafka liveness
321 if !liveness {
322 probe.UpdateStatusFromContext(ctx, "message-bus", probe.ServiceStatusNotReady)
323
324 if core.grpcServer != nil {
325 log.Info("kafka-manager-thread-set-server-notready")
326 }
327
328 // retry frequently while life is bad
Girish Kumar4d3887d2019-11-22 14:22:05 +0000329 timeout = notLiveProbeInterval
Scott Bakeree6a0872019-10-29 15:59:52 -0700330 } else {
331 probe.UpdateStatusFromContext(ctx, "message-bus", probe.ServiceStatusRunning)
332
333 if core.grpcServer != nil {
334 log.Info("kafka-manager-thread-set-server-ready")
335 }
336
337 // retry infrequently while life is good
Girish Kumar4d3887d2019-11-22 14:22:05 +0000338 timeout = liveProbeInterval
Scott Bakeree6a0872019-10-29 15:59:52 -0700339 }
340 if !timeoutTimer.Stop() {
341 <-timeoutTimer.C
342 }
343 case <-timeoutTimer.C:
344 log.Info("kafka-proxy-liveness-recheck")
345 // send the liveness probe in a goroutine; we don't want to deadlock ourselves as
346 // the liveness probe may wait (and block) writing to our channel.
347 go func() {
348 err := core.kmp.SendLiveness()
349 if err != nil {
350 // Catch possible error case if sending liveness after Sarama has been stopped.
351 log.Warnw("error-kafka-send-liveness", log.Fields{"error": err})
352 }
353 }()
354 }
355 }
khenaidoob9203542018-09-17 22:56:37 -0400356}
357
khenaidoob3244212019-08-27 14:32:27 -0400358// waitUntilKVStoreReachableOrMaxTries will wait until it can connect to a KV store or until maxtries has been reached
Girish Kumar4d3887d2019-11-22 14:22:05 +0000359func (core *Core) waitUntilKVStoreReachableOrMaxTries(ctx context.Context, maxRetries int, retryInterval time.Duration) error {
khenaidoob3244212019-08-27 14:32:27 -0400360 log.Infow("verifying-KV-store-connectivity", log.Fields{"host": core.config.KVStoreHost,
361 "port": core.config.KVStorePort, "retries": maxRetries, "retryInterval": retryInterval})
khenaidoob3244212019-08-27 14:32:27 -0400362 count := 0
363 for {
npujar467fe752020-01-16 20:17:45 +0530364 if !core.kvClient.IsConnectionUp(ctx) {
khenaidoob3244212019-08-27 14:32:27 -0400365 log.Info("KV-store-unreachable")
366 if maxRetries != -1 {
367 if count >= maxRetries {
368 return status.Error(codes.Unavailable, "kv store unreachable")
369 }
370 }
npujar1d86a522019-11-14 17:11:16 +0530371 count++
khenaidoob3244212019-08-27 14:32:27 -0400372 // Take a nap before retrying
Girish Kumar4d3887d2019-11-22 14:22:05 +0000373 time.Sleep(retryInterval)
khenaidoob3244212019-08-27 14:32:27 -0400374 log.Infow("retry-KV-store-connectivity", log.Fields{"retryCount": count, "maxRetries": maxRetries, "retryInterval": retryInterval})
375
376 } else {
377 break
378 }
379 }
380 log.Info("KV-store-reachable")
381 return nil
382}
383
npujar1d86a522019-11-14 17:11:16 +0530384func (core *Core) registerAdapterRequestHandlers(ctx context.Context, coreInstanceID string, dMgr *DeviceManager,
khenaidoo297cd252019-02-07 22:10:23 -0500385 ldMgr *LogicalDeviceManager, aMgr *AdapterManager, cdProxy *model.Proxy, ldProxy *model.Proxy,
khenaidoo54e0ddf2019-02-27 16:21:33 -0500386) error {
npujar1d86a522019-11-14 17:11:16 +0530387 requestProxy := NewAdapterRequestHandlerProxy(core, coreInstanceID, dMgr, ldMgr, aMgr, cdProxy, ldProxy,
khenaidoo297cd252019-02-07 22:10:23 -0500388 core.config.InCompetingMode, core.config.LongRunningRequestTimeout, core.config.DefaultRequestTimeout)
khenaidoob9203542018-09-17 22:56:37 -0400389
khenaidoo54e0ddf2019-02-27 16:21:33 -0500390 // Register the broadcast topic to handle any core-bound broadcast requests
391 if err := core.kmp.SubscribeWithRequestHandlerInterface(kafka.Topic{Name: core.config.CoreTopic}, requestProxy); err != nil {
392 log.Fatalw("Failed-registering-broadcast-handler", log.Fields{"topic": core.config.CoreTopic})
393 return err
394 }
395
Kent Hagermana6d0c362019-07-30 12:50:21 -0400396 // Register the core-pair topic to handle core-bound requests destined to the core pair
397 if err := core.kmp.SubscribeWithDefaultRequestHandler(kafka.Topic{Name: core.config.CorePairTopic}, kafka.OffsetNewest); err != nil {
398 log.Fatalw("Failed-registering-pair-handler", log.Fields{"topic": core.config.CorePairTopic})
399 return err
400 }
401
khenaidoo54e0ddf2019-02-27 16:21:33 -0500402 log.Info("request-handler-registered")
khenaidoob9203542018-09-17 22:56:37 -0400403 return nil
404}
405
406func (core *Core) startDeviceManager(ctx context.Context) {
khenaidoo21d51152019-02-01 13:48:37 -0500407 log.Info("DeviceManager-Starting...")
khenaidoo4d4802d2018-10-04 21:59:49 -0400408 core.deviceMgr.start(ctx, core.logicalDeviceMgr)
khenaidoo21d51152019-02-01 13:48:37 -0500409 log.Info("DeviceManager-Started")
khenaidoob9203542018-09-17 22:56:37 -0400410}
411
412func (core *Core) startLogicalDeviceManager(ctx context.Context) {
khenaidoo21d51152019-02-01 13:48:37 -0500413 log.Info("Logical-DeviceManager-Starting...")
khenaidoo4d4802d2018-10-04 21:59:49 -0400414 core.logicalDeviceMgr.start(ctx)
khenaidoo21d51152019-02-01 13:48:37 -0500415 log.Info("Logical-DeviceManager-Started")
khenaidoob9203542018-09-17 22:56:37 -0400416}
khenaidoo21d51152019-02-01 13:48:37 -0500417
418func (core *Core) startAdapterManager(ctx context.Context) {
419 log.Info("Adapter-Manager-Starting...")
Thomas Lee Se5a44012019-11-07 20:32:24 +0530420 err := core.adapterMgr.start(ctx)
421 if err != nil {
422 log.Fatalf("failed-to-start-adapter-manager: error %v ", err)
423 }
khenaidoo21d51152019-02-01 13:48:37 -0500424 log.Info("Adapter-Manager-Started")
William Kurkiandaa6bb22019-03-07 12:26:28 -0500425}
Girish Kumar4d3887d2019-11-22 14:22:05 +0000426
427/*
428* Thread to monitor kvstore Liveness (connection status)
429*
430* This function constantly monitors Liveness State of kvstore as reported
431* periodically by backend and updates the Status of kv-store service registered
432* with rw_core probe.
433*
434* If no liveness event has been seen within a timeout, then the thread will
435* perform a "liveness" check attempt, which will in turn trigger a liveness event on
436* the liveness channel, true or false depending on whether the attempt succeeded.
437*
438* The gRPC server in turn monitors the state of the readiness probe and will
439* start issuing UNAVAILABLE response while the probe is not ready.
440 */
441func (core *Core) monitorKvstoreLiveness(ctx context.Context) {
442 log.Info("start-monitoring-kvstore-liveness")
443
444 // Instruct backend to create Liveness channel for transporting state updates
445 livenessChannel := core.backend.EnableLivenessChannel()
446
447 log.Debug("enabled-kvstore-liveness-channel")
448
449 // Default state for kvstore is alive for rw_core
450 timeout := core.config.LiveProbeInterval
Scott Baker2d87ee32020-03-03 13:04:01 -0800451loop:
Girish Kumar4d3887d2019-11-22 14:22:05 +0000452 for {
453 timeoutTimer := time.NewTimer(timeout)
454 select {
455
456 case liveness := <-livenessChannel:
457 log.Debugw("received-liveness-change-notification", log.Fields{"liveness": liveness})
458
459 if !liveness {
460 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusNotReady)
461
462 if core.grpcServer != nil {
463 log.Info("kvstore-set-server-notready")
464 }
465
466 timeout = core.config.NotLiveProbeInterval
467
468 } else {
469 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusRunning)
470
471 if core.grpcServer != nil {
472 log.Info("kvstore-set-server-ready")
473 }
474
475 timeout = core.config.LiveProbeInterval
476 }
477
478 if !timeoutTimer.Stop() {
479 <-timeoutTimer.C
480 }
481
Scott Baker2d87ee32020-03-03 13:04:01 -0800482 case <-core.exitChannel:
483 break loop
484
Girish Kumar4d3887d2019-11-22 14:22:05 +0000485 case <-timeoutTimer.C:
486 log.Info("kvstore-perform-liveness-check-on-timeout")
487
488 // Trigger Liveness check if no liveness update received within the timeout period.
489 // The Liveness check will push Live state to same channel which this routine is
490 // reading and processing. This, do it asynchronously to avoid blocking for
491 // backend response and avoid any possibility of deadlock
npujar467fe752020-01-16 20:17:45 +0530492 go core.backend.PerformLivenessCheck(ctx)
Girish Kumar4d3887d2019-11-22 14:22:05 +0000493 }
494 }
495}