blob: 78965b74a4e7e9501ae273588f6d3e7c4de92a41 [file] [log] [blame]
Kent Hagerman2f0d0552020-04-23 17:28:52 -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 */
16
17package core
18
19import (
20 "context"
21 "errors"
22 "time"
23
Maninderdfadc982020-10-28 14:04:33 +053024 "github.com/opencord/voltha-lib-go/v4/pkg/db"
25 "github.com/opencord/voltha-lib-go/v4/pkg/db/kvstore"
26 "github.com/opencord/voltha-lib-go/v4/pkg/log"
27 "github.com/opencord/voltha-lib-go/v4/pkg/probe"
Kent Hagerman2f0d0552020-04-23 17:28:52 -040028 "google.golang.org/grpc/codes"
29 "google.golang.org/grpc/status"
30)
31
Rohan Agrawal31f21802020-06-12 05:38:46 +000032func newKVClient(ctx context.Context, storeType string, address string, timeout time.Duration) (kvstore.Client, error) {
33 logger.Infow(ctx, "kv-store-type", log.Fields{"store": storeType})
Kent Hagerman2f0d0552020-04-23 17:28:52 -040034 switch storeType {
35 case "consul":
Rohan Agrawal31f21802020-06-12 05:38:46 +000036 return kvstore.NewConsulClient(ctx, address, timeout)
Kent Hagerman2f0d0552020-04-23 17:28:52 -040037 case "etcd":
Rohan Agrawal31f21802020-06-12 05:38:46 +000038 return kvstore.NewEtcdClient(ctx, address, timeout, log.FatalLevel)
Kent Hagerman2f0d0552020-04-23 17:28:52 -040039 }
40 return nil, errors.New("unsupported-kv-store")
41}
42
43func stopKVClient(ctx context.Context, kvClient kvstore.Client) {
44 // Release all reservations
45 if err := kvClient.ReleaseAllReservations(ctx); err != nil {
Rohan Agrawal31f21802020-06-12 05:38:46 +000046 logger.Infow(ctx, "fail-to-release-all-reservations", log.Fields{"error": err})
Kent Hagerman2f0d0552020-04-23 17:28:52 -040047 }
48 // Close the DB connection
Rohan Agrawal31f21802020-06-12 05:38:46 +000049 kvClient.Close(ctx)
Kent Hagerman2f0d0552020-04-23 17:28:52 -040050}
51
52// waitUntilKVStoreReachableOrMaxTries will wait until it can connect to a KV store or until maxtries has been reached
53func waitUntilKVStoreReachableOrMaxTries(ctx context.Context, kvClient kvstore.Client, maxRetries int, retryInterval time.Duration) error {
Rohan Agrawal31f21802020-06-12 05:38:46 +000054 logger.Infow(ctx, "verifying-KV-store-connectivity", log.Fields{"retries": maxRetries, "retryInterval": retryInterval})
Kent Hagerman2f0d0552020-04-23 17:28:52 -040055 count := 0
56 for {
57 if !kvClient.IsConnectionUp(ctx) {
Rohan Agrawal31f21802020-06-12 05:38:46 +000058 logger.Info(ctx, "KV-store-unreachable")
Kent Hagerman2f0d0552020-04-23 17:28:52 -040059 if maxRetries != -1 {
60 if count >= maxRetries {
61 return status.Error(codes.Unavailable, "kv store unreachable")
62 }
63 }
64 count++
65
66 // Take a nap before retrying
67 select {
68 case <-ctx.Done():
69 //ctx canceled
70 return ctx.Err()
71 case <-time.After(retryInterval):
72 }
Rohan Agrawal31f21802020-06-12 05:38:46 +000073 logger.Infow(ctx, "retry-KV-store-connectivity", log.Fields{"retryCount": count, "maxRetries": maxRetries, "retryInterval": retryInterval})
Kent Hagerman2f0d0552020-04-23 17:28:52 -040074 } else {
75 break
76 }
77 }
78 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusRunning)
Rohan Agrawal31f21802020-06-12 05:38:46 +000079 logger.Info(ctx, "KV-store-reachable")
Kent Hagerman2f0d0552020-04-23 17:28:52 -040080 return nil
81}
82
83/*
84 * Thread to monitor kvstore Liveness (connection status)
85 *
86 * This function constantly monitors Liveness State of kvstore as reported
87 * periodically by backend and updates the Status of kv-store service registered
88 * with rw_core probe.
89 *
90 * If no liveness event has been seen within a timeout, then the thread will
91 * perform a "liveness" check attempt, which will in turn trigger a liveness event on
92 * the liveness channel, true or false depending on whether the attempt succeeded.
93 *
94 * The gRPC server in turn monitors the state of the readiness probe and will
95 * start issuing UNAVAILABLE response while the probe is not ready.
96 */
97func monitorKVStoreLiveness(ctx context.Context, backend *db.Backend, liveProbeInterval, notLiveProbeInterval time.Duration) {
Rohan Agrawal31f21802020-06-12 05:38:46 +000098 logger.Info(ctx, "start-monitoring-kvstore-liveness")
Kent Hagerman2f0d0552020-04-23 17:28:52 -040099
100 // Instruct backend to create Liveness channel for transporting state updates
Rohan Agrawal31f21802020-06-12 05:38:46 +0000101 livenessChannel := backend.EnableLivenessChannel(ctx)
Kent Hagerman2f0d0552020-04-23 17:28:52 -0400102
Rohan Agrawal31f21802020-06-12 05:38:46 +0000103 logger.Debug(ctx, "enabled-kvstore-liveness-channel")
Kent Hagerman2f0d0552020-04-23 17:28:52 -0400104
105 // Default state for kvstore is alive for rw_core
106 timeout := liveProbeInterval
107loop:
108 for {
109 timeoutTimer := time.NewTimer(timeout)
110 select {
111
112 case liveness := <-livenessChannel:
Rohan Agrawal31f21802020-06-12 05:38:46 +0000113 logger.Debugw(ctx, "received-liveness-change-notification", log.Fields{"liveness": liveness})
Kent Hagerman2f0d0552020-04-23 17:28:52 -0400114
115 if !liveness {
116 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusNotReady)
Rohan Agrawal31f21802020-06-12 05:38:46 +0000117 logger.Info(ctx, "kvstore-set-server-notready")
Kent Hagerman2f0d0552020-04-23 17:28:52 -0400118
119 timeout = notLiveProbeInterval
120
121 } else {
122 probe.UpdateStatusFromContext(ctx, "kv-store", probe.ServiceStatusRunning)
Rohan Agrawal31f21802020-06-12 05:38:46 +0000123 logger.Info(ctx, "kvstore-set-server-ready")
Kent Hagerman2f0d0552020-04-23 17:28:52 -0400124
125 timeout = liveProbeInterval
126 }
127
128 if !timeoutTimer.Stop() {
129 <-timeoutTimer.C
130 }
131
132 case <-ctx.Done():
133 break loop
134
135 case <-timeoutTimer.C:
Rohan Agrawal31f21802020-06-12 05:38:46 +0000136 logger.Info(ctx, "kvstore-perform-liveness-check-on-timeout")
Kent Hagerman2f0d0552020-04-23 17:28:52 -0400137
138 // Trigger Liveness check if no liveness update received within the timeout period.
139 // The Liveness check will push Live state to same channel which this routine is
140 // reading and processing. This, do it asynchronously to avoid blocking for
141 // backend response and avoid any possibility of deadlock
142 go backend.PerformLivenessCheck(ctx)
143 }
144 }
145}