// WatchForFinishedTestRuns simply watches the local cache for any test runs that have test data.
// It will periodically look at the table and send any present test data back to the server as a response.
// When the server has successfully received this data, it will send a task back to this specific agent
// to delete this row from the cache.
func WatchForFinishedTestRuns(cfg config.Config) {
var ac = cache.NewAgentCache(cfg)
agentUuid := ac.GetKeyValue("uuid")
for {
time.Sleep(5000 * time.Millisecond)
testruns, err := ac.GetFinishedTestRuns()
if err != nil {
log.Error("Problem retrieving finished test runs")
os.Exit(1)
}
for testUuid, testData := range testruns {
log.Debug("Found ripe testrun: ", testUuid)
var utdr = responses.UploadTestDataResponse{
TestUuid: testUuid,
TestData: testData,
}
utdr.AgentUuid = agentUuid
utdr.Type = "TestData" //TODO(mierdin): This is an extra step. Maybe a factory function for the task could help here?
var tc = comms.NewToDDComms(cfg)
tc.CommsPackage.SendResponse(utdr)
}
}
}
// Run contains the logic necessary to perform this task on the agent. This particular task
// will simply pass a key/value pair to the agent cache to be set
func (kvt KeyValueTask) Run() error {
var ac = cache.NewAgentCache(kvt.Config)
err := ac.SetKeyValue(kvt.Key, kvt.Value)
if err != nil {
return errors.New(fmt.Sprintf("KeyValueTask failed - %s:%s", kvt.Key, kvt.Value))
}
log.Info(fmt.Sprintf("KeyValueTask successful - %s:%s", kvt.Key, kvt.Value))
return nil
}
// WatchForGroup should be run as a goroutine, like other background services. This is because it will itself spawn a goroutine to
// listen for tasks that are sent to groups, and this goroutine can be restarted when group membership changes
func (rmq rabbitMQComms) WatchForGroup() {
var ac = cache.NewAgentCache(rmq.config)
// dereg is a channel that allows us to instruct the goroutine that's listening for tests to stop. This allows us to re-register to a new command
dereg := make(chan bool)
rereg:
group := ac.GetKeyValue("group")
// if the group is nothing, rewrite to "mull". This is being done for now so that we don't have to worry if the goroutine was started or not
// This way, it's always running, but if the agent is not in a group, it's listening on the "null" queue, which never has anything on it.
// This is a minor waste of resources on the agent, so TODO(mierdin): you should probably fix this at some point and figure out how to only run
// the goroutine when needed, but at the same time prevent the dereg channel from blocking unnecessarily in that case
//
// This will also handle the cases when the agent first starts up, and the key for this group isn't present in the cache, and therefore is "".
if group == "" {
group = "null"
}
go func() {
for {
err := rmq.ListenForGroupTasks(group, dereg)
if err != nil {
log.Warn("ListenForGroupTasks reported a failure. Trying again...")
}
}
}()
// Loop until the unackedGroup flag is set
for {
time.Sleep(2 * time.Second)
// The key "unackedGroup" stores a "true" or "false" to indicate that there has been a group change that we need to acknowledge (handle)
if ac.GetKeyValue("unackedGroup") == "true" {
// This will kill the underlying goroutine, and in effect stop listening to the old queue.
dereg <- true
// Finally, set the "unackedGroup" to indicate that we've acknowledged the group change, and go back to the "rereg" label
// to re-register onto the new group name
ac.SetKeyValue("unackedGroup", "false")
goto rereg
}
}
}
// Run contains the logic necessary to perform this task on the agent.
// This particular task will install, but not execute a testrun on this agent.
// The installation procedure will first run the referenced testlet in check mode
// to help ensure that the actual testrun execution can take place. If that
// succeeds, the testrun is installed in the agent cache.
func (itt InstallTestRunTask) Run() error {
if itt.Tr.Testlet == "" {
log.Error("Testlet parameter for this testrun is null")
return errors.New("Testlet parameter for this testrun is null")
}
// Generate path to testlet and make sure it exists.
testlet_path := fmt.Sprintf("%s/assets/testlets/%s", itt.Config.LocalResources.OptDir, itt.Tr.Testlet)
if _, err := os.Stat(testlet_path); os.IsNotExist(err) {
log.Errorf("Testlet %s does not exist on this agent", itt.Tr.Testlet)
return errors.New("Error installing testrun - testlet doesn't exist on this agent.")
}
// Run the testlet in check mode to verify that everything is okay to run this test
log.Debug("Running testlet in check mode: ", testlet_path)
cmd := exec.Command(testlet_path, "check")
// Stdout buffer
cmdOutput := &bytes.Buffer{}
// Attach buffer to command
cmd.Stdout = cmdOutput
// Execute collector
cmd.Run()
// This is probably the best cross-platform way to see if check mode passed.
if strings.Contains(string(cmdOutput.Bytes()), "Check mode PASSED") {
log.Debugf("Check mode for %s passed", testlet_path)
} else {
log.Error("Testlet returned an error during check mode: ", string(cmdOutput.Bytes()))
return errors.New("Testlet returned an error during check mode")
}
// Insert testrun into agent cache
var ac = cache.NewAgentCache(itt.Config)
err := ac.InsertTestRun(itt.Tr)
if err != nil {
log.Error(err)
return errors.New("Problem installing test run into agent cache")
}
return nil
}
// Run contains the logic necessary to perform this task on the agent.
func (sgt SetGroupTask) Run() error {
var ac = cache.NewAgentCache(sgt.Config)
// First, see what the current group is. If it matches what this task is instructing, we don't need to do anything.
if ac.GetKeyValue("group") != sgt.GroupName {
err := ac.SetKeyValue("group", sgt.GroupName)
if err != nil {
return errors.New(fmt.Sprintf("Failed to set keyvalue pair - %s:%s", "group", sgt.GroupName))
}
err = ac.SetKeyValue("unackedGroup", "true")
if err != nil {
return errors.New(fmt.Sprintf("Failed to set keyvalue pair - %s:%s", "unackedGroup", "true"))
}
} else {
log.Info("Already in the group being dictated by the server: ", sgt.GroupName)
}
return nil
}
func main() {
cfg := config.GetConfig(arg_config)
// Set up cache
var ac = cache.NewAgentCache(cfg)
ac.Init()
// Generate UUID
uuid := hostresources.GenerateUuid()
ac.SetKeyValue("uuid", uuid)
log.Infof("ToDD Agent Activated: %s", uuid)
// Start test data reporting service
go testing.WatchForFinishedTestRuns(cfg)
// Construct comms package
var tc = comms.NewToDDComms(cfg)
// Spawn goroutine to listen for tasks issued by server
go func() {
for {
err := tc.CommsPackage.ListenForTasks(uuid)
if err != nil {
log.Warn("ListenForTasks reported a failure. Trying again...")
}
}
}()
// Watch for changes to group membership
go tc.CommsPackage.WatchForGroup()
// Continually advertise agent status into message queue
for {
// Gather assets here as a map, and refer to a key in that map in the below struct
gatheredAssets := GetLocalAssets(cfg)
var defaultaddr string
if cfg.LocalResources.IPAddrOverride != "" {
defaultaddr = cfg.LocalResources.IPAddrOverride
} else {
defaultaddr = hostresources.GetIPOfInt(cfg.LocalResources.DefaultInterface).String()
}
// Create an AgentAdvert instance to represent this particular agent
me := defs.AgentAdvert{
Uuid: uuid,
DefaultAddr: defaultaddr,
FactCollectors: gatheredAssets["factcollectors"],
Testlets: gatheredAssets["testlets"],
Facts: facts.GetFacts(cfg),
LocalTime: time.Now().UTC(),
}
// Advertise this agent
err := tc.CommsPackage.AdvertiseAgent(me)
if err != nil {
log.Error("Failed to advertise agent after several retries")
}
time.Sleep(15 * time.Second) // TODO(moswalt): make configurable
}
}
// ListenForTasks is a method that recieves task notices from the server
func (rmq rabbitMQComms) ListenForTasks(uuid string) error {
// Connect to RabbitMQ with retry logic
conn, err := connectRabbitMQ(rmq.queueUrl)
if err != nil {
log.Error("(AdvertiseAgent) Failed to connect to RabbitMQ")
return err
}
defer conn.Close()
ch, err := conn.Channel()
if err != nil {
log.Error("Failed to open a channel")
os.Exit(1)
}
defer ch.Close()
_, err = ch.QueueDeclare(
uuid, // name
false, // durable
false, // delete when usused
false, // exclusive
false, // no-wait
nil, // arguments
)
if err != nil {
log.Error("Failed to declare a queue")
os.Exit(1)
}
msgs, err := ch.Consume(
uuid, // queue
uuid, // consumer
true, // auto-ack
false, // exclusive
false, // no-local
false, // no-wait
nil, // args
)
if err != nil {
log.Error("Failed to register a consumer")
os.Exit(1)
}
forever := make(chan bool)
go func() {
for d := range msgs {
// Unmarshal into BaseTaskMessage to determine type
var base_msg tasks.BaseTask
err = json.Unmarshal(d.Body, &base_msg)
// TODO(mierdin): Need to handle this error
log.Debugf("Agent task received: %s", d.Body)
// call agent task method based on type
switch base_msg.Type {
case "DownloadAsset":
downloadAssetTask := tasks.DownloadAssetTask{
HTTPClient: &http.Client{},
Fs: tasks.OsFS{},
Ios: tasks.IoSys{},
CollectorDir: fmt.Sprintf("%s/assets/factcollectors", rmq.config.LocalResources.OptDir),
TestletDir: fmt.Sprintf("%s/assets/testlets", rmq.config.LocalResources.OptDir),
}
err = json.Unmarshal(d.Body, &downloadAssetTask)
// TODO(mierdin): Need to handle this error
err = downloadAssetTask.Run()
if err != nil {
log.Warning("The KeyValue task failed to initialize")
}
case "KeyValue":
kv_task := tasks.KeyValueTask{
Config: rmq.config,
}
err = json.Unmarshal(d.Body, &kv_task)
// TODO(mierdin): Need to handle this error
err = kv_task.Run()
if err != nil {
log.Warning("The KeyValue task failed to initialize")
}
case "SetGroup":
sg_task := tasks.SetGroupTask{
Config: rmq.config,
}
err = json.Unmarshal(d.Body, &sg_task)
// TODO(mierdin): Need to handle this error
err = sg_task.Run()
if err != nil {
log.Warning("The SetGroup task failed to initialize")
}
case "DeleteTestData":
dtdt_task := tasks.DeleteTestDataTask{
Config: rmq.config,
}
err = json.Unmarshal(d.Body, &dtdt_task)
// TODO(mierdin): Need to handle this error
err = dtdt_task.Run()
if err != nil {
log.Warning("The DeleteTestData task failed to initialize")
}
case "InstallTestRun":
// Retrieve UUID
var ac = cache.NewAgentCache(rmq.config)
uuid := ac.GetKeyValue("uuid")
itr_task := tasks.InstallTestRunTask{
Config: rmq.config,
}
err = json.Unmarshal(d.Body, &itr_task)
// TODO(mierdin): Need to handle this error
var response responses.SetAgentStatusResponse
response.Type = "AgentStatus" //TODO(mierdin): This is an extra step. Maybe a factory function for the task could help here?
response.AgentUuid = uuid
response.TestUuid = itr_task.Tr.Uuid
err = itr_task.Run()
if err != nil {
log.Warning("The InstallTestRun task failed to initialize")
response.Status = "fail"
} else {
response.Status = "ready"
}
rmq.SendResponse(response)
case "ExecuteTestRun":
// Retrieve UUID
var ac = cache.NewAgentCache(rmq.config)
uuid := ac.GetKeyValue("uuid")
etr_task := tasks.ExecuteTestRunTask{
Config: rmq.config,
}
err = json.Unmarshal(d.Body, &etr_task)
// TODO(mierdin): Need to handle this error
// Send status that the testing has begun, right now.
response := responses.SetAgentStatusResponse{
TestUuid: etr_task.TestUuid,
Status: "testing",
}
response.AgentUuid = uuid // TODO(mierdin): Can't declare this in the literal, it's that embedding behavior again. Need to figure this out.
response.Type = "AgentStatus" //TODO(mierdin): This is an extra step. Maybe a factory function for the task could help here?
rmq.SendResponse(response)
err = etr_task.Run()
if err != nil {
log.Warning("The ExecuteTestRun task failed to initialize")
response.Status = "fail"
rmq.SendResponse(response)
}
default:
log.Errorf(fmt.Sprintf("Unexpected type value for received task: %s", base_msg.Type))
}
}
}()
log.Infof(" [*] Waiting for messages. To exit press CTRL+C")
<-forever
return nil
}