| 1 | // SPDX-License-Identifier: GPL-3.0-or-later |
| 2 | |
| 3 | package jobmgr |
| 4 | |
| 5 | import ( |
| 6 | "bytes" |
| 7 | "context" |
| 8 | "strings" |
| 9 | "sync" |
| 10 | "testing" |
| 11 | "time" |
| 12 | |
| 13 | "github.com/stretchr/testify/assert" |
| 14 | "github.com/stretchr/testify/require" |
| 15 | |
| 16 | "github.com/netdata/netdata/go/plugins/pkg/funcapi" |
| 17 | "github.com/netdata/netdata/go/plugins/plugin/agent/secrets/secretstore" |
| 18 | "github.com/netdata/netdata/go/plugins/plugin/framework/collectorapi" |
| 19 | "github.com/netdata/netdata/go/plugins/plugin/framework/confgroup" |
| 20 | "github.com/netdata/netdata/go/plugins/plugin/framework/functions" |
| 21 | "github.com/netdata/netdata/go/plugins/plugin/framework/vnodes" |
| 22 | ) |
| 23 | |
| 24 | func TestRunProcessConfGroups_ChannelCloseDoesNotSpin(t *testing.T) { |
| 25 | tests := map[string]struct { |
| 26 | closeInput bool |
| 27 | cancelBefore bool |
| 28 | wantExit bool |
| 29 | }{ |
| 30 | "closed channel exits loop": { |
| 31 | closeInput: true, |
| 32 | wantExit: true, |
| 33 | }, |
| 34 | "canceled context exits loop": { |
| 35 | cancelBefore: true, |
| 36 | wantExit: true, |
| 37 | }, |
| 38 | } |
| 39 | |
| 40 | for name, tc := range tests { |
| 41 | t.Run(name, func(t *testing.T) { |
| 42 | mgr := New(Config{PluginName: testPluginName}) |
| 43 | ctx, cancel := context.WithCancel(context.Background()) |
| 44 | t.Cleanup(cancel) |
| 45 | mgr.ctx = ctx |
| 46 | |
| 47 | in := make(chan []*confgroup.Group) |
| 48 | if tc.closeInput { |
| 49 | close(in) |
| 50 | } |
| 51 | if tc.cancelBefore { |
| 52 | cancel() |
| 53 | } |
| 54 | |
| 55 | done := make(chan struct{}) |
| 56 | go func() { |
| 57 | mgr.runProcessConfGroups(in) |
| 58 | close(done) |
| 59 | }() |
| 60 | |
| 61 | select { |
| 62 | case <-done: |
| 63 | assert.True(t, tc.wantExit) |
| 64 | case <-time.After(200 * time.Millisecond): |
| 65 | t.Fatal("runProcessConfGroups did not exit") |
| 66 | } |
| 67 | }) |
| 68 | } |
| 69 | } |
| 70 | |
| 71 | func TestRunNotifyRunningJobs_TickOutsideLock(t *testing.T) { |
| 72 | mgr := New(Config{PluginName: testPluginName}) |
| 73 | |
| 74 | ctx, cancel := context.WithCancel(context.Background()) |
| 75 | defer cancel() |
| 76 | mgr.ctx = ctx |
| 77 | |
| 78 | job := &lockProbeJob{ |
| 79 | fullName: "success_lockprobe", |
| 80 | moduleName: "success", |
| 81 | name: "lockprobe", |
| 82 | tickStarted: make(chan struct{}), |
| 83 | tickRelease: make(chan struct{}), |
| 84 | } |
| 85 | |
| 86 | mgr.runningJobs.lock() |
| 87 | mgr.runningJobs.add(job.FullName(), job) |
| 88 | mgr.runningJobs.unlock() |
| 89 | |
| 90 | done := make(chan struct{}) |
| 91 | go func() { |
| 92 | mgr.runNotifyRunningJobs() |
| 93 | close(done) |
| 94 | }() |
| 95 | |
| 96 | select { |
| 97 | case <-job.tickStarted: |
| 98 | case <-time.After(2 * time.Second): |
| 99 | t.Fatal("tick did not start") |
| 100 | } |
| 101 | |
| 102 | stopDone := make(chan struct{}) |
| 103 | go func() { |
| 104 | mgr.stopRunningJob(job.FullName()) |
| 105 | close(stopDone) |
| 106 | }() |
| 107 | |
| 108 | select { |
| 109 | case <-stopDone: |
| 110 | case <-time.After(300 * time.Millisecond): |
| 111 | t.Fatal("stopRunningJob blocked while Tick was in progress") |
| 112 | } |
| 113 | |
| 114 | close(job.tickRelease) |
| 115 | cancel() |
| 116 | |
| 117 | select { |
| 118 | case <-done: |
| 119 | case <-time.After(2 * time.Second): |
| 120 | t.Fatal("runNotifyRunningJobs did not stop") |
| 121 | } |
| 122 | } |
| 123 | |
| 124 | func TestRun_DoesNotRegisterModuleMethodsBeforeAnyJobStarts(t *testing.T) { |
| 125 | fnReg := &recordingFunctionRegistry{} |
| 126 | mgr := New(Config{PluginName: testPluginName, FnReg: fnReg}) |
| 127 | |
| 128 | mgr.modules = collectorapi.Registry{ |
| 129 | "mod": collectorapi.Creator{ |
| 130 | Methods: func() []funcapi.MethodConfig { |
| 131 | return []funcapi.MethodConfig{{ID: "a"}} |
| 132 | }, |
| 133 | }, |
| 134 | } |
| 135 | |
| 136 | ctx, cancel := context.WithCancel(context.Background()) |
| 137 | defer cancel() |
| 138 | |
| 139 | in := make(chan []*confgroup.Group) |
| 140 | done := make(chan struct{}) |
| 141 | go func() { |
| 142 | mgr.Run(ctx, in) |
| 143 | close(done) |
| 144 | }() |
| 145 | |
| 146 | waitCtx, waitCancel := context.WithTimeout(context.Background(), time.Second) |
| 147 | defer waitCancel() |
| 148 | require.True(t, mgr.WaitStarted(waitCtx), "manager did not report started") |
| 149 | |
| 150 | cancel() |
| 151 | close(in) |
| 152 | |
| 153 | select { |
| 154 | case <-done: |
| 155 | case <-time.After(2 * time.Second): |
| 156 | t.Fatal("manager did not stop after cancel") |
| 157 | } |
| 158 | |
| 159 | assert.Empty(t, fnReg.registeredNames(), "static methods must not be registered before first started job") |
| 160 | } |
| 161 | |
| 162 | func TestStartRunningJob_RegistersModuleMethodsOnFirstStartedJob(t *testing.T) { |
| 163 | fnReg := &recordingFunctionRegistry{} |
| 164 | mgr := New(Config{PluginName: testPluginName, FnReg: fnReg}) |
| 165 | creator := collectorapi.Creator{ |
| 166 | Methods: func() []funcapi.MethodConfig { |
| 167 | return []funcapi.MethodConfig{{ID: "a"}, {ID: "b"}} |
| 168 | }, |
| 169 | } |
| 170 | mgr.modules = collectorapi.Registry{"mod": creator} |
| 171 | mgr.funcCtl.RegisterModules(mgr.modules) |
| 172 | |
| 173 | job := &lockProbeJob{fullName: "mod_job1", moduleName: "mod", name: "job1"} |
| 174 | mgr.startRunningJob(job) |
| 175 | |
| 176 | assert.ElementsMatch(t, []string{"mod:a", "mod:b"}, fnReg.registeredNames()) |
| 177 | } |
| 178 | |
| 179 | func TestStartRunningJob_DoesNotReregisterModuleMethods(t *testing.T) { |
| 180 | fnReg := &recordingFunctionRegistry{} |
| 181 | mgr := New(Config{PluginName: testPluginName, FnReg: fnReg}) |
| 182 | creator := collectorapi.Creator{ |
| 183 | Methods: func() []funcapi.MethodConfig { |
| 184 | return []funcapi.MethodConfig{{ID: "a"}, {ID: "b"}} |
| 185 | }, |
| 186 | } |
| 187 | mgr.modules = collectorapi.Registry{"mod": creator} |
| 188 | mgr.funcCtl.RegisterModules(mgr.modules) |
| 189 | |
| 190 | job1 := &lockProbeJob{fullName: "mod_job1", moduleName: "mod", name: "job1"} |
| 191 | mgr.startRunningJob(job1) |
| 192 | |
| 193 | job2 := &lockProbeJob{fullName: "mod_job2", moduleName: "mod", name: "job2"} |
| 194 | mgr.startRunningJob(job2) |
| 195 | |
| 196 | registered := fnReg.registeredNames() |
| 197 | assert.Len(t, registered, 2) |
| 198 | assert.ElementsMatch(t, []string{"mod:a", "mod:b"}, registered) |
| 199 | } |
| 200 | |
| 201 | func TestRun_RegistersDyncfgConfigPrefixes(t *testing.T) { |
| 202 | fnReg := &recordingFunctionRegistry{} |
| 203 | mgr := New(Config{PluginName: testPluginName, FnReg: fnReg}) |
| 204 | |
| 205 | ctx, cancel := context.WithCancel(context.Background()) |
| 206 | defer cancel() |
| 207 | |
| 208 | in := make(chan []*confgroup.Group) |
| 209 | done := make(chan struct{}) |
| 210 | go func() { |
| 211 | mgr.Run(ctx, in) |
| 212 | close(done) |
| 213 | }() |
| 214 | |
| 215 | waitCtx, waitCancel := context.WithTimeout(context.Background(), time.Second) |
| 216 | defer waitCancel() |
| 217 | require.True(t, mgr.WaitStarted(waitCtx), "manager did not report started") |
| 218 | |
| 219 | cancel() |
| 220 | close(in) |
| 221 | |
| 222 | select { |
| 223 | case <-done: |
| 224 | case <-time.After(2 * time.Second): |
| 225 | t.Fatal("manager did not stop after cancel") |
| 226 | } |
| 227 | |
| 228 | assert.Equal(t, []registeredPrefix{ |
| 229 | {name: "config", prefix: mgr.dyncfgVnodePrefixValue()}, |
| 230 | {name: "config", prefix: mgr.dyncfgSecretStorePrefixValue()}, |
| 231 | {name: "config", prefix: mgr.dyncfgCollectorPrefixValue()}, |
| 232 | }, fnReg.registeredPrefixes()) |
| 233 | } |
| 234 | |
| 235 | func TestRun_PublishesVnodesAndSecretstoresBeforeCollectorTemplates(t *testing.T) { |
| 236 | fnReg := &recordingFunctionRegistry{} |
| 237 | var buf bytes.Buffer |
| 238 | |
| 239 | mgr := New(Config{ |
| 240 | PluginName: testPluginName, |
| 241 | FnReg: fnReg, |
| 242 | Out: &buf, |
| 243 | Modules: collectorapi.Registry{ |
| 244 | "mod": collectorapi.Creator{}, |
| 245 | }, |
| 246 | }) |
| 247 | |
| 248 | ctx, cancel := context.WithCancel(context.Background()) |
| 249 | defer cancel() |
| 250 | |
| 251 | in := make(chan []*confgroup.Group) |
| 252 | done := make(chan struct{}) |
| 253 | go func() { |
| 254 | mgr.Run(ctx, in) |
| 255 | close(done) |
| 256 | }() |
| 257 | |
| 258 | waitCtx, waitCancel := context.WithTimeout(context.Background(), time.Second) |
| 259 | defer waitCancel() |
| 260 | require.True(t, mgr.WaitStarted(waitCtx), "manager did not report started") |
| 261 | |
| 262 | cancel() |
| 263 | close(in) |
| 264 | |
| 265 | select { |
| 266 | case <-done: |
| 267 | case <-time.After(2 * time.Second): |
| 268 | t.Fatal("manager did not stop after cancel") |
| 269 | } |
| 270 | |
| 271 | output := buf.String() |
| 272 | |
| 273 | vnodeIdx := strings.Index(output, "CONFIG "+mgr.dyncfgVnodePrefixValue()+" create accepted template /collectors/"+testPluginName+"/Vnodes") |
| 274 | secretIdx := strings.Index(output, "CONFIG "+mgr.dyncfgSecretStoreID(string(secretstore.KindVault))+" create accepted template /collectors/"+testPluginName+"/SecretStores") |
| 275 | collectorIdx := strings.Index(output, "CONFIG "+mgr.dyncfgModID("mod")+" create accepted template /collectors/"+testPluginName+"/Jobs") |
| 276 | |
| 277 | require.NotEqual(t, -1, vnodeIdx, "vnode template publication not found") |
| 278 | require.NotEqual(t, -1, secretIdx, "secretstore template publication not found") |
| 279 | require.NotEqual(t, -1, collectorIdx, "collector template publication not found") |
| 280 | assert.Less(t, vnodeIdx, collectorIdx, "vnode publication must happen before collector template publication") |
| 281 | assert.Less(t, secretIdx, collectorIdx, "secretstore publication must happen before collector template publication") |
| 282 | } |
| 283 | |
| 284 | type lockProbeJob struct { |
| 285 | fullName string |
| 286 | moduleName string |
| 287 | name string |
| 288 | |
| 289 | tickOnce sync.Once |
| 290 | stopOnce sync.Once |
| 291 | tickStarted chan struct{} |
| 292 | tickRelease chan struct{} |
| 293 | } |
| 294 | |
| 295 | func (j *lockProbeJob) FullName() string { return j.fullName } |
| 296 | func (j *lockProbeJob) ModuleName() string { return j.moduleName } |
| 297 | func (j *lockProbeJob) Name() string { return j.name } |
| 298 | func (j *lockProbeJob) Collector() any { return nil } |
| 299 | func (j *lockProbeJob) Start() {} |
| 300 | func (j *lockProbeJob) Stop() { j.stopOnce.Do(func() {}) } |
| 301 | func (j *lockProbeJob) Tick(_ int) { |
| 302 | j.tickOnce.Do(func() { |
| 303 | close(j.tickStarted) |
| 304 | <-j.tickRelease |
| 305 | }) |
| 306 | } |
| 307 | func (j *lockProbeJob) AutoDetection() error { return nil } |
| 308 | func (j *lockProbeJob) AutoDetectionEvery() int { return 0 } |
| 309 | func (j *lockProbeJob) RetryAutoDetection() bool { return false } |
| 310 | func (j *lockProbeJob) Cleanup() {} |
| 311 | func (j *lockProbeJob) IsRunning() bool { return true } |
| 312 | func (j *lockProbeJob) Panicked() bool { return false } |
| 313 | func (j *lockProbeJob) Vnode() vnodes.VirtualNode { return vnodes.VirtualNode{} } |
| 314 | func (j *lockProbeJob) UpdateVnode(_ *vnodes.VirtualNode) {} |
| 315 | |
| 316 | type recordingFunctionRegistry struct { |
| 317 | mu sync.Mutex |
| 318 | registered []string |
| 319 | prefixes []registeredPrefix |
| 320 | } |
| 321 | |
| 322 | func (r *recordingFunctionRegistry) Register(name string, _ func(functions.Function)) { |
| 323 | r.mu.Lock() |
| 324 | r.registered = append(r.registered, name) |
| 325 | r.mu.Unlock() |
| 326 | } |
| 327 | |
| 328 | func (r *recordingFunctionRegistry) Unregister(string) {} |
| 329 | func (r *recordingFunctionRegistry) RegisterPrefix(name, prefix string, _ func(functions.Function)) { |
| 330 | r.mu.Lock() |
| 331 | r.prefixes = append(r.prefixes, registeredPrefix{name: name, prefix: prefix}) |
| 332 | r.mu.Unlock() |
| 333 | } |
| 334 | func (r *recordingFunctionRegistry) UnregisterPrefix(string, string) {} |
| 335 | |
| 336 | func (r *recordingFunctionRegistry) registeredNames() []string { |
| 337 | r.mu.Lock() |
| 338 | defer r.mu.Unlock() |
| 339 | out := make([]string, len(r.registered)) |
| 340 | copy(out, r.registered) |
| 341 | return out |
| 342 | } |
| 343 | |
| 344 | func (r *recordingFunctionRegistry) registeredPrefixes() []registeredPrefix { |
| 345 | r.mu.Lock() |
| 346 | defer r.mu.Unlock() |
| 347 | out := make([]registeredPrefix, len(r.prefixes)) |
| 348 | copy(out, r.prefixes) |
| 349 | return out |
| 350 | } |
| 351 | |
| 352 | type registeredPrefix struct { |
| 353 | name string |
| 354 | prefix string |
| 355 | } |