fix(watchdog): improve logging context in keepalive error handling
This commit is contained in:
@@ -17,12 +17,13 @@ func StartKeepAlive() error {
|
|||||||
|
|
||||||
if err := isProcessAlive(); err != nil {
|
if err := isProcessAlive(); err != nil {
|
||||||
global.Is.WatchdogConnected = false
|
global.Is.WatchdogConnected = false
|
||||||
postLog.Warning(fmt.Sprintf("[watchdog] keepalive check failed: %v", err))
|
postLog.Warning(fmt.Sprintf("[watchdog.StartKeepAlive] keepalive check failed: %v", err))
|
||||||
|
|
||||||
_ = Disconnect()
|
_ = Disconnect()
|
||||||
|
|
||||||
var lastErr error
|
var lastErr error
|
||||||
for i := 0; i < 5; i++ {
|
for i := 0; i < 5; i++ {
|
||||||
|
KillWatchdogProcess()
|
||||||
if err := ensureWatchdogProcess(); err != nil {
|
if err := ensureWatchdogProcess(); err != nil {
|
||||||
lastErr = err
|
lastErr = err
|
||||||
time.Sleep(500 * time.Millisecond)
|
time.Sleep(500 * time.Millisecond)
|
||||||
@@ -31,7 +32,7 @@ func StartKeepAlive() error {
|
|||||||
|
|
||||||
if err = Init(); err == nil && Connect() {
|
if err = Init(); err == nil && Connect() {
|
||||||
global.Is.WatchdogConnected = true
|
global.Is.WatchdogConnected = true
|
||||||
postLog.Info("[watchdog] successfully reconnected to watchdog")
|
postLog.Info("[watchdog.StartKeepAlive] successfully reconnected to watchdog")
|
||||||
lastErr = nil
|
lastErr = nil
|
||||||
break
|
break
|
||||||
}
|
}
|
||||||
@@ -41,7 +42,7 @@ func StartKeepAlive() error {
|
|||||||
}
|
}
|
||||||
|
|
||||||
if lastErr != nil {
|
if lastErr != nil {
|
||||||
postLog.Error(fmt.Sprintf("[watchdog] failed to recover watchdog connection: %v", lastErr))
|
postLog.Error(fmt.Sprintf("[watchdog.StartKeepAlive] failed to recover watchdog connection: %v", lastErr))
|
||||||
return fmt.Errorf("watchdog recovery failed after 5 attempts: %w", lastErr)
|
return fmt.Errorf("watchdog recovery failed after 5 attempts: %w", lastErr)
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|||||||
Reference in New Issue
Block a user