Set timeout and re-dial on any failures.

zfs_replication_exporter.go

@@ -151,16 +151,23 @@ func main() {
 		Auth: ams,
 		// TODO(wathiede); use FixedHostKey?
 		HostKeyCallback: ssh.InsecureIgnoreHostKey(),
-	}
+		Timeout:         5 * time.Second,
-	c, err := ssh.Dial("tcp", *host, config)
-	if err != nil {
-		glog.Exitf("Error dialing %q: %v", *host, err)
 	}
 
 	go func() {
 		for {
+			var c *ssh.Client
+			if c == nil {
+				var err error
+				c, err = ssh.Dial("tcp", *host, config)
+				if err != nil {
+					glog.Errorf("Error dialing %q: %v", *host, err)
+				}
+			}
 			if err := updateMetrics(c); err != nil {
 				glog.Errorf("Failed to update metrics: %v", err)
+				c.Close()
+				c = nil
 			}
 			time.Sleep(*refreshInterval)
 		}