func (etcd *ETCDClusterRunner) stop(nuke bool) {
etcd.mutex.Lock()
defer etcd.mutex.Unlock()
if etcd.running {
for i := 0; i < etcd.numNodes; i++ {
ginkgomon.Interrupt(etcd.etcdProcesses[i], 5*time.Second)
if nuke {
etcd.nukeArtifacts(i)
}
}
etcd.markAsStopped()
}
}
func (runner *AgentRunner) Stop() {
runner.mutex.Lock()
defer runner.mutex.Unlock()
if !runner.running {
return
}
ginkgomon.Interrupt(runner.consulProcess, 5*time.Second)
os.RemoveAll(runner.dataDir)
os.RemoveAll(runner.configDir)
runner.consulProcess = nil
runner.running = false
}
func (cr *ClusterRunner) Stop() {
cr.mutex.Lock()
defer cr.mutex.Unlock()
if !cr.running {
return
}
for i := 0; i < cr.numNodes; i++ {
ginkgomon.Interrupt(cr.consulProcesses[i], 5*time.Second)
}
os.RemoveAll(cr.dataDir)
os.RemoveAll(cr.configDir)
cr.consulProcesses = nil
cr.running = false
}
Expect(err).NotTo(HaveOccurred())
service, ok := services[registration.ID]
Expect(ok).To(BeTrue())
Expect(*service).To(Equal(api.AgentService{
ID: registration.ID,
Service: registration.Name,
Tags: registration.Tags,
Port: registration.Port,
Address: registration.Address,
}))
})
})
Context("when signalled", func() {
It("deregisters the given service before exiting", func() {
ginkgomon.Interrupt(registrationProcess)
services, err := consulClient.Agent().Services()
Expect(err).NotTo(HaveOccurred())
Expect(services).ToNot(HaveKey(registration.ID))
})
})
})
var _ = Describe("Service Registration Unit Tests", func() {
var (
client *fakes.FakeClient
agent *fakes.FakeAgent
logger lager.Logger
clock *fakeclock.FakeClock
registration *api.AgentServiceRegistration
Eventually(presenceProcess.Wait()).ShouldNot(Receive())
clock.WaitForWatcherAndIncrement(6 * time.Second)
Eventually(logger).Should(Say("recreating-session"))
consulRunner.Start()
consulRunner.WaitUntilReady()
clock.WaitForWatcherAndIncrement(6 * time.Second)
Eventually(logger).Should(Say("succeeded-recreating-session"))
Eventually(presenceProcess.Ready()).Should(BeClosed())
})
})
Context("and the process is shutting down", func() {
It("releases the presence and exits", func() {
ginkgomon.Interrupt(presenceProcess)
Eventually(presenceProcess.Wait()).Should(Receive(BeNil()))
_, err := getPresenceValue()
Expect(err).To(Equal(consuladapter.NewKeyNotFoundError(presenceKey)))
})
})
Context("and consul goes through a period of instability", func() {
var serveFiveHundreds chan struct{}
var fakeConsul *httptest.Server
BeforeEach(func() {
serveFiveHundreds = make(chan struct{}, 8)
consulClusterURL, err := url.Parse(consulRunner.URL())
Expect(err).NotTo(HaveOccurred())
return httpClient.Do(req)
}
BeforeEach(func() {
nsyncPort = 8888 + GinkgoParallelNode()
nsyncURL := fmt.Sprintf("http://127.0.0.1:%d", nsyncPort)
requestGenerator = rata.NewRequestGenerator(nsyncURL, nsync.Routes)
httpClient = http.DefaultClient
runner := newNSyncRunner(fmt.Sprintf("127.0.0.1:%d", nsyncPort))
process = ginkgomon.Invoke(runner)
})
AfterEach(func() {
ginkgomon.Interrupt(process, exitDuration)
})
Describe("Desire an app", func() {
BeforeEach(func() {
response, err = requestDesireWithInstances(3)
})
It("desires the app from the bbs", func() {
Expect(err).NotTo(HaveOccurred())
Expect(response.StatusCode).To(Equal(http.StatusAccepted))
Eventually(func() ([]*models.DesiredLRP, error) {
return bbsClient.DesiredLRPs(models.DesiredLRPFilter{})
}, 10).Should(HaveLen(1))
desiredLrps, _ := bbsClient.DesiredLRPs(models.DesiredLRPFilter{})
newRouteMessage := json.RawMessage([]byte(`[{"hostnames":["route-1"],"port":8080}]`))
var pipelineDBFactory db.PipelineDBFactory
BeforeEach(func() {
dbLogger := lagertest.NewTestLogger("test")
postgresRunner.Truncate()
dbConn = postgresRunner.Open()
dbListener = pq.NewListener(postgresRunner.DataSourceName(), time.Second, time.Minute, nil)
bus := db.NewNotificationsBus(dbListener, dbConn)
sqlDB = db.NewSQL(dbLogger, dbConn, bus)
pipelineDBFactory = db.NewPipelineDBFactory(dbLogger, dbConn, bus, sqlDB)
atcProcess, atcPort = startATC(atcBin, 1)
})
AfterEach(func() {
ginkgomon.Interrupt(atcProcess)
Expect(dbConn.Close()).To(Succeed())
Expect(dbListener.Close()).To(Succeed())
})
Describe("viewing a list of builds", func() {
var page *agouti.Page
var pipelineDB db.PipelineDB
BeforeEach(func() {
var err error
page, err = agoutiDriver.NewPage()
Expect(err).NotTo(HaveOccurred())
})
BeforeEach(func() {
dbLogger := lagertest.NewTestLogger("test")
postgresRunner.Truncate()
dbConn = db.Wrap(postgresRunner.Open())
dbListener = pq.NewListener(postgresRunner.DataSourceName(), time.Second, time.Minute, nil)
bus := db.NewNotificationsBus(dbListener, dbConn)
sqlDB = db.NewSQL(dbLogger, dbConn, bus)
atcOneProcess, atcOnePort = startATC(atcBin, 1, true, BASIC_AUTH)
atcTwoProcess, atcTwoPort = startATC(atcBin, 2, true, BASIC_AUTH)
})
AfterEach(func() {
ginkgomon.Interrupt(atcOneProcess)
ginkgomon.Interrupt(atcTwoProcess)
Expect(dbConn.Close()).To(Succeed())
Expect(dbListener.Close()).To(Succeed())
})
Describe("Pipes", func() {
var client *http.Client
BeforeEach(func() {
client = &http.Client{
Transport: &http.Transport{},
}
})
func (m *MetronRunner) Stop() {
ginkgomon.Interrupt(m.Process, 2)
}
const cell1 = "cell-id-1"
const cell2 = "cell-id-2"
Context("when there is a single cell", func() {
var maintainers ifrit.Process
BeforeEach(func() {
Expect(serviceClient.Cells(logger)).To(HaveLen(0))
maintainers = ifrit.Invoke(grouper.NewParallel(os.Interrupt, grouper.Members{
{cell1, serviceClient.NewCellPresenceRunner(logger, newCellPresence(cell1), locket.RetryInterval, locket.LockTTL)},
{cell2, serviceClient.NewCellPresenceRunner(logger, newCellPresence(cell2), locket.RetryInterval, locket.LockTTL)},
}))
})
AfterEach(func() {
ginkgomon.Interrupt(maintainers)
})
It("returns only one cell", func() {
Eventually(func() (models.CellSet, error) { return serviceClient.Cells(logger) }).Should(HaveLen(2))
Expect(serviceClient.Cells(logger)).To(HaveKey(cell1))
Expect(serviceClient.Cells(logger)).To(HaveKey(cell2))
})
})
})
})
func newCellPresence(cellID string) *models.CellPresence {
presence := models.NewCellPresence(
cellID,
"cell.example.com",
registrations = make(chan *nats.Msg, 1)
natsClient.Subscribe("router.register", func(msg *nats.Msg) {
registrations <- msg
})
})
It("starts announcing its location again", func() {
Eventually(registrations).Should(Receive())
})
})
})
Context("when a server is sent SIGINT after the hearbeat has started", func() {
JustBeforeEach(func() {
Eventually(receptorRunner).Should(gbytes.Say("nats-heartbeat.started"))
ginkgomon.Interrupt(receptorProcess)
})
Context("and NATS is accessible", func() {
var unregistrations chan *nats.Msg
BeforeEach(func() {
unregistrations = make(chan *nats.Msg, 1)
natsClient.Subscribe("router.unregister", func(msg *nats.Msg) {
unregistrations <- msg
})
})
It("broadcasts an unregister message", func() {
Eventually(unregistrations).Should(Receive())
})
)
tsaRunner := ginkgomon.New(ginkgomon.Config{
Command: tsaCommand,
Name: "tsa",
StartCheck: "tsa.listening",
AnsiColorCode: "32m",
})
tsaProcess = ginkgomon.Invoke(tsaRunner)
})
AfterEach(func() {
atcServer.Close()
gardenServer.Stop()
ginkgomon.Interrupt(tsaProcess)
})
Describe("SSHing", func() {
var sshSess *gexec.Session
var sshStdin io.Writer
var sshArgv []string
BeforeEach(func() {
sshArgv = []string{
"127.0.0.1",
"-p", strconv.Itoa(tsaPort),
"-o", "UserKnownHostsFile=" + userKnownHostsFile,
}
})
Expect(delta).To(BeEquivalentTo(1))
})
})
It("cleans up exiting connections when killing the BBS", func(done Done) {
var err error
eventSource, err = client.SubscribeToEvents(logger)
Expect(err).NotTo(HaveOccurred())
go func() {
_, err := eventSource.Next()
Expect(err).To(HaveOccurred())
close(done)
}()
ginkgomon.Interrupt(bbsProcess)
})
})
func primeEventStream(eventChannel chan models.Event, eventType string, primer func(), cleanup func()) {
primer()
PRIMING:
for {
select {
case <-eventChannel:
break PRIMING
case <-time.After(50 * time.Millisecond):
cleanup()
primer()
}
Describe("pinging the server", func() {
var pingErr error
Context("when Garden responds to ping", func() {
JustBeforeEach(func() {
pingErr = executorClient.Ping()
})
It("does not return an error", func() {
Expect(pingErr).NotTo(HaveOccurred())
})
})
Context("when Garden returns an error", func() {
JustBeforeEach(func() {
ginkgomon.Interrupt(gardenProcess)
pingErr = executorClient.Ping()
})
AfterEach(func() {
gardenProcess = ginkgomon.Invoke(componentMaker.GardenLinux())
})
It("should return an error", func() {
Expect(pingErr).To(HaveOccurred())
Expect(pingErr.Error()).To(ContainSubstring("connection refused"))
})
})
})
Describe("getting the total resources", func() {
etcdStoreClient = etcd.NewStoreClient(nil)
})
Context("but SQL does not have a version", func() {
BeforeEach(func() {
fakeSQLDB.VersionReturns(nil, models.ErrResourceNotFound)
})
It("fetches the version from etcd", func() {
Eventually(fakeSQLDB.VersionCallCount).Should(Equal(1))
Consistently(fakeSQLDB.VersionCallCount).Should(Equal(1))
Eventually(fakeETCDDB.VersionCallCount).Should(Equal(1))
Consistently(fakeETCDDB.VersionCallCount).Should(Equal(1))
ginkgomon.Interrupt(migrationProcess)
Eventually(migrationProcess.Wait()).Should(Receive(BeNil()))
})
// cross-db migration
Context("but etcd does", func() {
var fakeMigrationToSQL *migrationfakes.FakeMigration
BeforeEach(func() {
fakeMigrationToSQL = &migrationfakes.FakeMigration{}
fakeMigrationToSQL.VersionReturns(102)
fakeMigrationToSQL.RequiresSQLReturns(true)
dbVersion.CurrentVersion = 99
dbVersion.TargetVersion = 99
fakeMigration.VersionReturns(100)
It("unregisters from etcd when the process exits", func() {
routingAPIRunner := testrunner.New(routingAPIBinPath, routingAPIArgs)
proc := ifrit.Invoke(routingAPIRunner)
getRoutes := func() string {
routesPath := fmt.Sprintf("%s/v2/keys/routes", etcdUrl)
resp, err := http.Get(routesPath)
Expect(err).ToNot(HaveOccurred())
body, err := ioutil.ReadAll(resp.Body)
Expect(err).ToNot(HaveOccurred())
return string(body)
}
Eventually(getRoutes).Should(ContainSubstring("api.example.com/routing"))
ginkgomon.Interrupt(proc)
Eventually(getRoutes).ShouldNot(ContainSubstring("api.example.com/routing"))
Eventually(routingAPIRunner.ExitCode()).Should(Equal(0))
})
Context("when router groups endpoint is invoked", func() {
var proc ifrit.Process
BeforeEach(func() {
routingAPIRunner := testrunner.New(routingAPIBinPath, routingAPIArgs)
proc = ifrit.Invoke(routingAPIRunner)
})
AfterEach(func() {
ginkgomon.Interrupt(proc)
func stopMetron() {
ginkgomon.Interrupt(metronProcess)
}
err = natsClient.Publish(msg.Reply, response)
Expect(err).NotTo(HaveOccurred())
})
})
Context("when the emitter is running", func() {
var emitter ifrit.Process
BeforeEach(func() {
runner := createEmitterRunner("emitter1")
runner.StartCheck = "emitter1.started"
emitter = ginkgomon.Invoke(runner)
})
AfterEach(func() {
ginkgomon.Interrupt(emitter, emitterInterruptTimeout)
})
Context("and an lrp with routes is desired", func() {
BeforeEach(func() {
err := bbsClient.DesireLRP(desiredLRP)
Expect(err).NotTo(HaveOccurred())
})
Context("and an instance starts", func() {
BeforeEach(func() {
err := bbsClient.StartActualLRP(&lrpKey, &instanceKey, &netInfo)
Expect(err).NotTo(HaveOccurred())
})
It("emits its routes immediately", func() {
BeforeEach(func() {
message1 := marshalMessage(createContainerMetric("some-process-guid", 0, 3.0, 1024, 2048, 0))
message2 := marshalMessage(createContainerMetric("some-process-guid", 1, 4.0, 1024, 2048, 0))
message3 := marshalMessage(createContainerMetric("some-process-guid", 2, 5.0, 1024, 2048, 0))
messages := map[string][][]byte{}
messages["some-log-guid"] = [][]byte{message1, message2, message3}
handler := NewHttpHandler(messages)
httpServer := http_server.New(trafficControllerAddress, handler)
trafficControllerProcess = ifrit.Invoke(sigmon.New(httpServer))
Expect(trafficControllerProcess.Ready()).To(BeClosed())
})
AfterEach(func() {
ginkgomon.Interrupt(trafficControllerProcess)
})
It("reports the state of the given process guid's instances", func() {
getLRPStats, err := requestGenerator.CreateRequest(
tps.LRPStats,
rata.Params{"guid": "some-process-guid"},
nil,
)
Expect(err).NotTo(HaveOccurred())
getLRPStats.Header.Add("Authorization", "I can do this.")
response, err := httpClient.Do(getLRPStats)
Expect(err).NotTo(HaveOccurred())
Expect(response.StatusCode).To(Equal(http.StatusOK))
process = ifrit.Invoke(
converger_process.New(
fakeBBSServiceClient,
fakeBBSClient,
logger,
fakeClock,
convergeRepeatInterval,
kickTaskDuration,
expirePendingTaskDuration,
expireCompletedTaskDuration,
),
)
})
AfterEach(func() {
ginkgomon.Interrupt(process)
Eventually(process.Wait()).Should(Receive())
})
Describe("converging over time", func() {
It("converges tasks, LRPs, and auctions when the lock is periodically reestablished", func() {
fakeClock.Increment(convergeRepeatInterval + aBit)
Eventually(fakeBBSClient.ConvergeTasksCallCount, aBit).Should(Equal(1))
Eventually(fakeBBSClient.ConvergeLRPsCallCount, aBit).Should(Equal(1))
actualKickTaskDuration, actualExpirePendingTaskDuration, actualExpireCompletedTaskDuration := fakeBBSClient.ConvergeTasksArgsForCall(0)
Expect(actualKickTaskDuration).To(Equal(kickTaskDuration))
Expect(actualExpirePendingTaskDuration).To(Equal(expirePendingTaskDuration))
Expect(actualExpireCompletedTaskDuration).To(Equal(expireCompletedTaskDuration))
serviceClient.NewCellPresenceRunnerReturns(fakeHeartbeater)
config = maintain.Config{
CellID: "cell-id",
RepAddress: "1.2.3.4",
Zone: "az1",
RetryInterval: 1 * time.Second,
RootFSProviders: []string{"provider-1", "provider-2"},
}
maintainer = maintain.New(config, fakeClient, serviceClient, logger, clock)
})
AfterEach(func() {
logger.Info("test-complete-signaling-maintainer-to-stop")
close(pingErrors)
ginkgomon.Interrupt(maintainProcess)
})
It("pings the executor", func() {
pingErrors <- nil
maintainProcess = ginkgomon.Invoke(maintainer)
Expect(fakeClient.PingCallCount()).To(Equal(1))
})
Context("when pinging the executor fails", func() {
It("keeps pinging until it succeeds, then starts heartbeating the executor's presence", func() {
maintainProcess = ifrit.Background(maintainer)
ready := maintainProcess.Ready()
for i := 1; i <= 4; i++ {
clock.Increment(1 * time.Second)
Describe("AuctioneerAddress", func() {
Context("when able to get an auctioneer presence", func() {
var heartbeater ifrit.Process
var presence auctioneer.Presence
BeforeEach(func() {
presence = auctioneer.NewPresence("auctioneer-id", "auctioneer.example.com")
auctioneerLock, err := serviceClient.NewAuctioneerLockRunner(logger, presence, 100*time.Millisecond, 10*time.Second)
Expect(err).NotTo(HaveOccurred())
heartbeater = ginkgomon.Invoke(auctioneerLock)
})
AfterEach(func() {
ginkgomon.Interrupt(heartbeater)
})
It("returns the address", func() {
address, err := serviceClient.CurrentAuctioneerAddress()
Expect(err).NotTo(HaveOccurred())
Expect(address).To(Equal(presence.AuctioneerAddress))
})
})
Context("when unable to get any auctioneer presences", func() {
It("returns ErrServiceUnavailable", func() {
_, err := serviceClient.CurrentAuctioneerAddress()
Expect(err).To(HaveOccurred())
})
})
AfterEach(func() {
consulRunner.Start()
consulRunner.WaitUntilReady()
})
It("loses the lock and exits", func() {
var err error
Eventually(lockProcess.Wait()).Should(Receive(&err))
Expect(err).To(Equal(locket.ErrLockLost))
Expect(sender.GetValue(lockHeldMetricName).Value).To(Equal(float64(0)))
})
})
Context("and the process is shutting down", func() {
It("releases the lock and exits", func() {
ginkgomon.Interrupt(lockProcess)
Eventually(lockProcess.Wait()).Should(Receive(BeNil()))
_, err := getLockValue()
Expect(err).To(Equal(consuladapter.NewKeyNotFoundError(lockKey)))
})
})
Context("and consul goes through a period of instability", func() {
var serveFiveHundreds chan struct{}
var fakeConsul *httptest.Server
BeforeEach(func() {
serveFiveHundreds = make(chan struct{}, 4)
consulClusterURL, err := url.Parse(consulRunner.URL())
Expect(err).NotTo(HaveOccurred())
w.Write(payload)
}),
)
})
AfterEach(func() {
defer fakeCC.Close()
})
Describe("when the CC polling interval elapses", func() {
JustBeforeEach(func() {
process = startBulker(true)
})
AfterEach(func() {
ginkgomon.Interrupt(process, interruptTimeout)
})
Context("once the state has been synced with CC", func() {
Context("lrps", func() {
BeforeEach(func() {
schedulingInfoResponse := models.DesiredLRPSchedulingInfosResponse{
Error: nil,
DesiredLrpSchedulingInfos: []*models.DesiredLRPSchedulingInfo{
{
DesiredLRPKey: models.DesiredLRPKey{ // perfect. love it. keep it.
ProcessGuid: "process-guid-1",
Domain: cc_messages.AppLRPDomain,
},
Annotation: "1.1",
},
It("POSTs to the CC that the application has crashed", func() {
Eventually(ready, 5*time.Second).Should(BeClosed())
})
})
Context("when the watcher loses the lock", func() {
BeforeEach(func() {
watcher, _ = startWatcher(true)
})
JustBeforeEach(func() {
consulRunner.Reset()
})
AfterEach(func() {
ginkgomon.Interrupt(watcher, 5)
})
It("exits with an error", func() {
Eventually(watcher.Wait(), 5).Should(Receive(HaveOccurred()))
})
})
Context("when the watcher initially does not have the lock", func() {
var runner *ginkgomon.Runner
var otherSession *consuladapter.Session
BeforeEach(func() {
otherSession = consulRunner.NewSession("other-Session")
err := otherSession.AcquireLock(locket.LockSchemaPath(watcherLockName), []byte("something-else"))
Expect(err).NotTo(HaveOccurred())
