func main() {
todd_version := "0.0.1"
cfg := config.GetConfig(arg_config)
// Start serving collectors and testlets, and retrieve map of names and hashes
assets := serveAssets(cfg)
// Perform database initialization tasks
tdb, err := db.NewToddDB(cfg)
if err != nil {
log.Fatalf("Error setting up database: %v\n", err)
}
if err := tdb.Init(); err != nil {
log.Fatalf("Error initializing database: %v\n", err)
}
// Initialize API
var tapi toddapi.ToDDApi
go func() {
log.Fatal(tapi.Start(cfg))
}()
// Start listening for agent advertisements
var tc = comms.NewToDDComms(cfg)
go tc.CommsPackage.ListenForAgent(assets)
// Kick off group calculation in background
go func() {
for {
log.Info("Beginning group calculation")
grouping.CalculateGroups(cfg)
time.Sleep(time.Second * time.Duration(cfg.Grouping.Interval))
}
}()
log.Infof("ToDD server v%s. Press any key to exit...\n", todd_version)
// Sssh, sssh, only dreams now....
for {
time.Sleep(time.Second * 10)
}
}
// CalculateGroups is a function designed to ingest a list of group objects, a collection of agents, and
// return a map that contains the resulting group for each agent UUID.
func CalculateGroups(cfg config.Config) {
tdb, err := db.NewToddDB(cfg)
if err != nil {
log.Fatalf("Error connecting to DB: %v", err)
}
// Retrieve all currently active agents
agents, err := tdb.GetAgents()
if err != nil {
log.Fatalf("Error retrieving agents: %v", err)
}
// Retrieve all objects with type "group"
group_objs, err := tdb.GetObjects("group")
if err != nil {
log.Fatalf("Error retrieving groups: %v", err)
}
// Cast retrieved slice of ToddObject interfaces to actual GroupObjects
groups := make([]objects.GroupObject, len(group_objs))
for i, gobj := range group_objs {
groups[i] = gobj.(objects.GroupObject)
}
// groupmap contains the uuid-to-groupname mappings to be used for test runs
groupmap := map[string]string{}
// This slice will hold all of the agents that are sad because they didn't get into a group
var lonelyAgents []defs.AgentAdvert
next:
for x := range agents {
for i := range groups {
// See if this agent is in this group
if isInGroup(groups[i].Spec.Matches, agents[x].Facts) {
// Insert this group name ("Label") into groupmap under the key of the UUID for the agent that belongs to it
log.Debugf("Agent %s is in group %s\n", agents[x].Uuid, groups[i].Label)
groupmap[agents[x].Uuid] = groups[i].Label
continue next
}
}
// The "continue next" should prohibit all agents that have a group from getting to this point,
// so the only ones left do not have a group.
lonelyAgents = append(lonelyAgents, agents[x])
}
// Write results to database
err = tdb.SetGroupMap(groupmap)
if err != nil {
log.Fatalf("Error setting group map: %v", err)
}
// Send notifications to each agent to let them know what group they're in, so they can cache it
var tc = comms.NewToDDComms(cfg)
for uuid, groupName := range groupmap {
setGroupTask := tasks.SetGroupTask{
GroupName: groupName,
}
setGroupTask.Type = "SetGroup" //TODO(mierdin): Apparently this is necessary because inner type promotion doesn't apply for struct literals?
tc.CommsPackage.SendTask(uuid, setGroupTask)
}
// need to send a message to all agents that weren't in groupmap to set their group to nothing
for x := range lonelyAgents {
setGroupTask := tasks.SetGroupTask{
GroupName: "",
}
setGroupTask.Type = "SetGroup" //TODO(mierdin): Apparently this is necessary because inner type promotion doesn't apply for struct literals?
tc.CommsPackage.SendTask(lonelyAgents[x].Uuid, setGroupTask)
}
}
// ListenForResponses listens for responses from an agent
func (rmq rabbitMQComms) ListenForResponses(stopListeningForResponses *chan bool) {
queueName := "agentresponses"
conn, err := amqp.Dial(rmq.queueUrl)
if err != nil {
log.Error(err)
log.Error("Failed to connect to RabbitMQ")
os.Exit(1)
}
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(
queueName, // 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(
queueName, // queue
queueName, // 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)
}
tdb, err := db.NewToddDB(rmq.config) // TODO(vcabbage): Consider moving this into the rabbitMQComms struct
if err != nil {
log.Error("Failed to connect to DB")
os.Exit(1)
}
go func() {
for d := range msgs {
// Unmarshal into BaseResponse to determine type
var base_msg responses.BaseResponse
err = json.Unmarshal(d.Body, &base_msg)
// TODO(mierdin): Need to handle this error
log.Debugf("Agent response received: %s", d.Body)
// call agent response method based on type
switch base_msg.Type {
case "AgentStatus":
var sasr responses.SetAgentStatusResponse
err = json.Unmarshal(d.Body, &sasr)
// TODO(mierdin): Need to handle this error
log.Debugf("Agent %s is '%s' regarding test %s. Writing to DB.", sasr.AgentUuid, sasr.Status, sasr.TestUuid)
err := tdb.SetAgentTestStatus(sasr.TestUuid, sasr.AgentUuid, sasr.Status)
if err != nil {
log.Errorf("Error writing agent status to DB: %v", err)
}
case "TestData":
var utdr responses.UploadTestDataResponse
err = json.Unmarshal(d.Body, &utdr)
// TODO(mierdin): Need to handle this error
err = tdb.SetAgentTestData(utdr.TestUuid, utdr.AgentUuid, utdr.TestData)
// TODO(mierdin): Need to handle this error
// Send task to the agent that says to delete the entry
var dtdt tasks.DeleteTestDataTask
dtdt.Type = "DeleteTestData" //TODO(mierdin): This is an extra step. Maybe a factory function for the task could help here?
dtdt.TestUuid = utdr.TestUuid
rmq.SendTask(utdr.AgentUuid, dtdt)
// Finally, set the status for this agent in the test to "finished"
err := tdb.SetAgentTestStatus(dtdt.TestUuid, utdr.AgentUuid, "finished")
if err != nil {
log.Errorf("Error writing agent status to DB: %v", err)
}
default:
log.Errorf(fmt.Sprintf("Unexpected type value for received response: %s", base_msg.Type))
}
}
}()
log.Infof(" [*] Waiting for messages. To exit press CTRL+C")
<-*stopListeningForResponses
}
// ListenForAgent will listen on the message queue for new agent advertisements.
// It is meant to be run as a goroutine
func (rmq rabbitMQComms) ListenForAgent(assets map[string]map[string]string) {
// TODO(mierdin): does func param need to be a pointer?
conn, err := amqp.Dial(rmq.queueUrl)
if err != nil {
log.Error(err)
log.Error("Failed to connect to RabbitMQ")
os.Exit(1)
}
defer conn.Close()
ch, err := conn.Channel()
if err != nil {
log.Error("Failed to open a channel")
os.Exit(1)
}
defer ch.Close()
q, err := ch.QueueDeclare(
"agentadvert", // name
false, // durable
false, // delete when unused
false, // exclusive
false, // no-wait
nil, // arguments
)
if err != nil {
log.Error("Failed to declare a queue")
os.Exit(1)
}
msgs, err := ch.Consume(
q.Name, // queue
"agentadvert", // 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 {
log.Debugf("Agent advertisement recieved: %s", d.Body)
var agent defs.AgentAdvert
err = json.Unmarshal(d.Body, &agent)
// TODO(mierdin): Need to handle this error
// assetList is a slice that will contain any URLs that need to be sent to an
// agent as a response to an incorrect or incomplete list of assets
var assetList []string
// assets is the asset map from the SERVER's perspective
for asset_type, asset_hashes := range assets {
var agentAssets map[string]string
// agentAssets is the asset map from the AGENT's perspective
if asset_type == "factcollectors" {
agentAssets = agent.FactCollectors
} else if asset_type == "testlets" {
agentAssets = agent.Testlets
}
for name, hash := range asset_hashes {
// See if the hashes match (a missing asset will also result in False)
if agentAssets[name] != hash {
// hashes do not match, so we need to append the asset download URL to the remediate list
var default_ip string
if rmq.config.LocalResources.IPAddrOverride != "" {
default_ip = rmq.config.LocalResources.IPAddrOverride
} else {
default_ip = hostresources.GetIPOfInt(rmq.config.LocalResources.DefaultInterface).String()
}
asset_url := fmt.Sprintf("http://%s:%s/%s/%s", default_ip, rmq.config.Assets.Port, asset_type, name)
assetList = append(assetList, asset_url)
}
}
}
// Asset list is empty, so we can continue
if len(assetList) == 0 {
var tdb, _ = db.NewToddDB(rmq.config)
tdb.SetAgent(agent)
// This block of code checked that the agent time was within a certain range of the server time. If there was a large enough
// time skew, the agent advertisement would be rejected.
// I have disabled this for now - My plan was to use this to synchronize testrun execution amongst agents, but I have
// a solution to that for now. May revisit this later.
//
// Determine difference between server and agent time
// t1 := time.Now()
// var diff float64
// diff = t1.Sub(agent.LocalTime).Seconds()
//
// // If Agent is within half a second of server time, add insert to database
// if diff < 0.5 && diff > -0.5 {
// } else {
// // We don't want to register an agent if there is a severe time difference,
// // in order to ensure continuity during tests. So, just print log message.
// log.Warn("Agent time not within boundaries.")
// }
} else {
log.Warnf("Agent %s did not have the required asset files. This advertisement is ignored.", agent.Uuid)
var task tasks.DownloadAssetTask
task.Type = "DownloadAsset" //TODO(mierdin): This is an extra step. Maybe a factory function for the task could help here?
task.Assets = assetList
rmq.SendTask(agent.Uuid, task)
}
}
}()
log.Infof(" [*] Waiting for messages. To exit press CTRL+C")
<-forever
}