blob: 984036ce8486363345edb8628443a71d663f218c [file] [log] [blame]
David K. Bainbridgeb4a9ab02019-09-20 15:12:16 -07001/*
2 * Copyright 2019-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 */
16package probe
17
18import (
19 "context"
20 "fmt"
21 "github.com/opencord/voltha-go/common/log"
22 "net/http"
23 "sync"
24)
25
26// ProbeContextKey used to fetch the Probe instance from a context
27type ProbeContextKeyType string
28
29// ServiceStatus typed values for service status
30type ServiceStatus int
31
32const (
33 // ServiceStatusUnknown initial state of services
34 ServiceStatusUnknown ServiceStatus = iota
35
36 // ServiceStatusPreparing to optionally be used for prep, such as connecting
37 ServiceStatusPreparing
38
39 // ServiceStatusPrepared to optionally be used when prep is complete, but before run
40 ServiceStatusPrepared
41
42 // ServiceStatusRunning service is functional
43 ServiceStatusRunning
44
45 // ServiceStatusStopped service has stopped, but not because of error
46 ServiceStatusStopped
47
48 // ServiceStatusFailed service has stopped because of an error
49 ServiceStatusFailed
50)
51
52const (
53 // ProbeContextKey value of context key to fetch probe
54 ProbeContextKey = ProbeContextKeyType("status-update-probe")
55)
56
57// String convert ServiceStatus values to strings
58func (s ServiceStatus) String() string {
59 switch s {
60 default:
61 fallthrough
62 case ServiceStatusUnknown:
63 return "Unknown"
64 case ServiceStatusPreparing:
65 return "Preparing"
66 case ServiceStatusPrepared:
67 return "Prepared"
68 case ServiceStatusRunning:
69 return "Running"
70 case ServiceStatusStopped:
71 return "Stopped"
72 case ServiceStatusFailed:
73 return "Failed"
74 }
75}
76
77// ServiceStatusUpdate status update event
78type ServiceStatusUpdate struct {
79 Name string
80 Status ServiceStatus
81}
82
83// Probe reciever on which to implement probe capabilities
84type Probe struct {
85 readyFunc func(map[string]ServiceStatus) bool
86 healthFunc func(map[string]ServiceStatus) bool
87
88 mutex sync.RWMutex
89 status map[string]ServiceStatus
90 isReady bool
91 isHealthy bool
92}
93
94// WithReadyFunc override the default ready calculation function
95func (p *Probe) WithReadyFunc(readyFunc func(map[string]ServiceStatus) bool) *Probe {
96 p.readyFunc = readyFunc
97 return p
98}
99
100// WithHealthFunc override the default health calculation function
101func (p *Probe) WithHealthFunc(healthFunc func(map[string]ServiceStatus) bool) *Probe {
102 p.healthFunc = healthFunc
103 return p
104}
105
106// RegisterService register one or more service names with the probe, status will be track against service name
107func (p *Probe) RegisterService(names ...string) {
108 p.mutex.Lock()
109 defer p.mutex.Unlock()
110 if p.status == nil {
111 p.status = make(map[string]ServiceStatus)
112 }
113 for _, name := range names {
114 if _, ok := p.status[name]; !ok {
115 p.status[name] = ServiceStatusUnknown
116 log.Debugw("probe-service-registered", log.Fields{"service-name": name})
117 }
118 }
119}
120
121// UpdateStatus utility function to send a service update to the probe
122func (p *Probe) UpdateStatus(name string, status ServiceStatus) {
123 p.mutex.Lock()
124 defer p.mutex.Unlock()
125 if p.status == nil {
126 p.status = make(map[string]ServiceStatus)
127 }
128 p.status[name] = status
129 if p.readyFunc != nil {
130 p.isReady = p.readyFunc(p.status)
131 } else {
132 p.isReady = defaultReadyFunc(p.status)
133 }
134
135 if p.healthFunc != nil {
136 p.isHealthy = p.healthFunc(p.status)
137 } else {
138 p.isHealthy = defaultHealthFunc(p.status)
139 }
140 log.Debugw("probe-service-status-updated",
141 log.Fields{
142 "service-name": name,
143 "status": status.String(),
144 "ready": p.isReady,
145 "health": p.isHealthy,
146 })
147}
148
149// UpdateStatusFromContext a convenience function to pull the Probe reference from the
150// Context, if it exists, and then calling UpdateStatus on that Probe reference. If Context
151// is nil or if a Probe reference is not associated with the ProbeContextKey then nothing
152// happens
153func UpdateStatusFromContext(ctx context.Context, name string, status ServiceStatus) {
154 if ctx != nil {
155 if value := ctx.Value(ProbeContextKey); value != nil {
156 if p, ok := value.(*Probe); ok {
157 p.UpdateStatus(name, status)
158 }
159 }
160 }
161}
162
163// pulled out to a function to help better enable unit testing
164func (p *Probe) readzFunc(w http.ResponseWriter, req *http.Request) {
165 p.mutex.RLock()
166 defer p.mutex.RUnlock()
167 if p.isReady {
168 w.WriteHeader(http.StatusOK)
169 } else {
170 w.WriteHeader(http.StatusTeapot)
171 }
172}
173func (p *Probe) healthzFunc(w http.ResponseWriter, req *http.Request) {
174 p.mutex.RLock()
175 defer p.mutex.RUnlock()
176 if p.isHealthy {
177 w.WriteHeader(http.StatusOK)
178 } else {
179 w.WriteHeader(http.StatusTeapot)
180 }
181}
182func (p *Probe) detailzFunc(w http.ResponseWriter, req *http.Request) {
183 p.mutex.RLock()
184 defer p.mutex.RUnlock()
185 w.Header().Set("Content-Type", "application/json")
186 w.Write([]byte("{"))
187 comma := ""
188 for c, s := range p.status {
189 w.Write([]byte(fmt.Sprintf("%s\"%s\": \"%s\"", comma, c, s.String())))
190 comma = ", "
191 }
192 w.Write([]byte("}"))
193 w.WriteHeader(http.StatusOK)
194
195}
196
197// ListenAndServe implements 3 HTTP endpoints on the given port for healthz, readz, and detailz. Returns only on error
198func (p *Probe) ListenAndServe(port int) {
199 mux := http.NewServeMux()
200
201 // Returns the result of the readyFunc calculation
202 mux.HandleFunc("/readz", p.readzFunc)
203
204 // Returns the result of the healthFunc calculation
205 mux.HandleFunc("/healthz", p.healthzFunc)
206
207 // Returns the details of the services and their status as JSON
208 mux.HandleFunc("/detailz", p.detailzFunc)
209 s := &http.Server{
210 Addr: fmt.Sprintf(":%d", port),
211 Handler: mux,
212 }
213 log.Fatal(s.ListenAndServe())
214}
215
216// defaultReadyFunc if all services are running then ready, else not
217func defaultReadyFunc(services map[string]ServiceStatus) bool {
218 if len(services) == 0 {
219 return false
220 }
221 for _, status := range services {
222 if status != ServiceStatusRunning {
223 return false
224 }
225 }
226 return true
227}
228
229// defaultHealthFunc if no service is stopped or failed, then healthy, else not.
230// service is start as unknown, so they are considered healthy
231func defaultHealthFunc(services map[string]ServiceStatus) bool {
232 if len(services) == 0 {
233 return false
234 }
235 for _, status := range services {
236 if status == ServiceStatusStopped || status == ServiceStatusFailed {
237 return false
238 }
239 }
240 return true
241}