moncheck - add timeout option

Before the timeout for checks was static. With this change it is finally
an option to configure.

1 files changed, 12 insertions, 8 deletions

diff --git a/cmd/moncheck/main.go b/cmd/moncheck/main.go
index fbc2b57..cc05594 100644
--- a/cmd/moncheck/main.go
+++ b/cmd/moncheck/main.go
@@ -26,8 +26,9 @@ var (
 
 type (
 	Config struct {
-		DB   string `json:"db"`
-		Wait string `json:"wait"`
+		DB      string `json:"db"`
+		Timeout string `json:"timeout"`
+		Wait    string `json:"wait"`
 	}
 
 	States []int
@@ -40,7 +41,7 @@ func main() {
 	if err != nil {
 		log.Fatalf("could not read config: %s", err)
 	}
-	config := Config{}
+	config := Config{Timeout: "30s", Wait: "30s"}
 	if err := json.Unmarshal(raw, &config); err != nil {
 		log.Fatalf("could not parse config: %s", err)
 	}
@@ -49,6 +50,10 @@ func main() {
 	if err != nil {
 		log.Fatalf("could not parse wait duration: %s", err)
 	}
+	timeout, err := time.ParseDuration(config.Timeout)
+	if err != nil {
+		log.Fatalf("could not parse timeout: %s", err)
+	}
 
 	db, err := sql.Open("postgres", config.DB)
 	if err != nil {
@@ -56,14 +61,14 @@ func main() {
 	}
 
 	for i := 0; i < 25; i++ {
-		go check(i, db, waitDuration)
+		go check(i, db, waitDuration, timeout)
 	}
 	wg := sync.WaitGroup{}
 	wg.Add(1)
 	wg.Wait()
 }
 
-func check(thread int, db *sql.DB, waitDuration time.Duration) {
+func check(thread int, db *sql.DB, waitDuration, timeout time.Duration) {
 	for {
 		tx, err := db.Begin()
 		if err != nil {
@@ -101,20 +106,19 @@ func check(thread int, db *sql.DB, waitDuration time.Duration) {
 			tx.Rollback()
 			continue
 		}
-		ctx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
+		ctx, cancel := context.WithTimeout(context.Background(), timeout)
 		cmd := exec.CommandContext(ctx, cmdLine[0], cmdLine[1:]...)
 		output := bytes.NewBuffer([]byte{})
 		cmd.Stdout = output
 		cmd.Stderr = output
 		err = cmd.Run()
 		if err != nil && ctx.Err() == context.DeadlineExceeded {
-			log.Printf("[%d] check took too long: %s", id, err)
 			cancel()
 			// TODO which state to choose?
 			// TODO add notification handler
 			// TODO all this casting should be done better
 			states.Add(99)
-			output.Write([]byte(ctx.Err().Error()))
+			fmt.Fprintf(output, "check took longer than %s", timeout)
 		} else if err != nil {
 			cancel()
 			status, ok := cmd.ProcessState.Sys().(syscall.WaitStatus)