func StartHeartbeats(localIp string, ttl time.Duration, config *config.Config, storeAdapter storeadapter.StoreAdapter, logger *gosteno.Logger) (stopChan chan (chan bool)) {
if len(config.EtcdUrls) == 0 {
return
}
if storeAdapter == nil {
panic("store adapter is nil")
}
logger.Debugf("Starting Health Status Updates to Store: /healthstatus/doppler/%s/%s/%d", config.Zone, config.JobName, config.Index)
status, stopChan, err := storeAdapter.MaintainNode(storeadapter.StoreNode{
Key: fmt.Sprintf("/healthstatus/doppler/%s/%s/%d", config.Zone, config.JobName, config.Index),
Value: []byte(localIp),
TTL: uint64(ttl.Seconds()),
})
if err != nil {
panic(err)
}
go func() {
for stat := range status {
logger.Debugf("Health updates channel pushed %v at time %v", stat, time.Now())
}
}()
return stopChan
}
func Announce(localIP string, ttl time.Duration, config *config.Config, storeAdapter storeadapter.StoreAdapter, logger *gosteno.Logger) chan (chan bool) {
dopplerMetaBytes, err := buildDopplerMeta(localIP, config)
if err != nil {
panic(err)
}
key := fmt.Sprintf("%s/%s/%s/%d", META_ROOT, config.Zone, config.JobName, config.Index)
logger.Debugf("Starting Health Status Updates to Store: %s", key)
node := storeadapter.StoreNode{
Key: key,
Value: dopplerMetaBytes,
TTL: uint64(ttl.Seconds()),
}
// Call to create to make sure node is created before we return
storeAdapter.Create(node)
status, stopChan, err := storeAdapter.MaintainNode(node)
if err != nil {
panic(err)
}
// The status channel needs to be drained to maintain the node within the etcd cluster
go func() {
for stat := range status {
logger.Debugf("Health updates channel pushed %v at time %v", stat, time.Now())
}
}()
return stopChan
}
func AnnounceLegacy(localIP string, ttl time.Duration, config *config.Config, storeAdapter storeadapter.StoreAdapter, logger *gosteno.Logger) chan (chan bool) {
key := fmt.Sprintf("%s/%s/%s/%d", LEGACY_ROOT, config.Zone, config.JobName, config.Index)
status, stopChan, err := storeAdapter.MaintainNode(storeadapter.StoreNode{
Key: key,
Value: []byte(localIP),
TTL: uint64(ttl.Seconds()),
})
if err != nil {
panic(err)
}
// The status channel needs to be drained to maintain the node within the etcd cluster
go func() {
for stat := range status {
logger.Debugf("Health updates channel pushed %v at time %v", stat, time.Now())
}
}()
return stopChan
}
func Daemonize(
component string,
callback func() error,
period time.Duration,
timeout time.Duration,
logger logger.Logger,
adapter storeadapter.StoreAdapter,
) error {
logger.Info("Acquiring lock for " + component)
lock := storeadapter.StoreNode{
Key: "/hm/locks/" + component,
TTL: 10,
}
lostLockChannel, releaseLockChannel, err := adapter.MaintainNode(lock)
if err != nil {
logger.Info(fmt.Sprintf("Failed to acquire lock: %s", err))
return err
}
go func() {
<-lostLockChannel
logger.Error("Lost the lock", errors.New("Lock the lock"))
os.Exit(197)
}()
logger.Info("Acquired lock for " + component)
logger.Info(fmt.Sprintf("Running Daemon every %d seconds with a timeout of %d", int(period.Seconds()), int(timeout.Seconds())))
for {
afterChan := time.After(period)
timeoutChan := time.After(timeout)
errorChan := make(chan error, 1)
t := time.Now()
go func() {
errorChan <- callback()
}()
select {
case err := <-errorChan:
logger.Info("Daemonize Time", map[string]string{
"Component": component,
"Duration": fmt.Sprintf("%.4f", time.Since(t).Seconds()),
})
if err != nil {
logger.Error("Daemon returned an error. Continuining...", err)
}
case <-timeoutChan:
released := make(chan bool)
releaseLockChannel <- released
<-released
return errors.New("Daemon timed out. Aborting!")
}
<-afterChan
}
return nil
}
