Hello @q2683252 thank you for filing this issue and welcome to the Go project!

So perhaps since there isn't Delve support on s390x, it would be great to know the Process ID (PID) of each process and then while it is seemingly stuck waiting to copy, getting a stacktrace of what is happening e.g.
Perhaps you can modify your cmd.CombinedOutput to:
a) invoke cmd.Start() with your own buffer
b) Print out the Process ID (PID) of the process
c) Invoke cmd.Wait()
d) While any of the processes is stuck, send it a SIGQUIT e.g. kill -s QUIT <PID>
and examine that trace to see what your program is doing -- if your program while produce that core dump

The output of each of those can then show you what's going on so

--- orig.go	2019-08-18 23:30:51.000000000 -0600
+++ main.go	2019-08-18 23:30:14.000000000 -0600
@@ -13,8 +13,21 @@
 		bin := filepath.Join(cur_dir, name)
 		conf := filepath.Join(cur_dir, fmt.Sprintf("../config/%s.yml", name))
 		cmd := exec.Command(bin, "-c", conf)
+		buf := new(bytes.Buffer)
+		cmd.Stdout = buf
+		cmd.Stderr = buf
 		oss_logger.Infof("Athena start %v", cmd)
-		cmdOutput, err := cmd.CombinedOutput()
+		if err := cmd.Start(); err != nil {
+			oss_logger.Errorf("%s service start error: %v", name, err)
+			return
+		}
+
+		// This is the PID of the subcommand that we shall
+		// send that "SIQUIT" if it stalls for a long time.
+		oss_logger.Info("%s service PID: %d\n", cmd.Process.Pid)
+
+		err = cmd.Wait()
+		cmdOutput := buf.String()
 		if err != nil {
 			oss_logger.Errorf("%s service out put %s", name, cmdOutput)
 			oss_logger.Errorf("%s service failed:%s,service restart", name, err.Error())

so once you have the PID printed out

kill -s QUIT XXXX

In the meantime I shall ping an s390x expert @mundaym about this bug.

Hello @q2683252 thank you for filing this issue and welcome to the Go project!

So perhaps since there isn't Delve support on s390x, it would be great to know the Process ID (PID) of each process and then while it is seemingly stuck waiting to copy, getting a stacktrace of what is happening e.g.
Perhaps you can modify your cmd.CombinedOutput to:
a) invoke cmd.Start() with your own buffer
b) Print out the Process ID (PID) of the process
c) Invoke cmd.Wait()
d) While any of the processes is stuck, send it a SIGQUIT e.g. kill -s QUIT <PID>
and examine that trace to see what your program is doing -- if your program while produce that core dump

The output of each of those can then show you what's going on so

--- orig.go	2019-08-18 23:30:51.000000000 -0600
+++ main.go	2019-08-18 23:30:14.000000000 -0600
@@ -13,8 +13,21 @@
 		bin := filepath.Join(cur_dir, name)
 		conf := filepath.Join(cur_dir, fmt.Sprintf("../config/%s.yml", name))
 		cmd := exec.Command(bin, "-c", conf)
+		buf := new(bytes.Buffer)
+		cmd.Stdout = buf
+		cmd.Stderr = buf
 		oss_logger.Infof("Athena start %v", cmd)
-		cmdOutput, err := cmd.CombinedOutput()
+		if err := cmd.Start(); err != nil {
+			oss_logger.Errorf("%s service start error: %v", name, err)
+			return
+		}
+
+		// This is the PID of the subcommand that we shall
+		// send that "SIQUIT" if it stalls for a long time.
+		oss_logger.Info("%s service PID: %d\n", cmd.Process.Pid)
+
+		err = cmd.Wait()
+		cmdOutput := buf.String()
 		if err != nil {
 			oss_logger.Errorf("%s service out put %s", name, cmdOutput)
 			oss_logger.Errorf("%s service failed:%s,service restart", name, err.Error())

so once you have the PID printed out

kill -s QUIT XXXX

In the meantime I shall ping an s390x expert @mundaym about this bug.
Thank you for your kindly reply.
It's not bound to happen everytime it got exited.
But what should I do if I have the pid?

Thank you for your kindly reply.
It's not bound to happen everytime it got exited.
But what should I do if I have the pid?

Once you have the PID, if the process is stuck, then send it a signal to cause it to produce a core dump and that will show you what's stopping progress on the forked process. I showed that in my initial comment part d) #33704 (comment) but also here again

kill -s QUIT XXXX

where XXXX is the printed out PID of the process that's stuck.

Thank you for your kindly reply.
It's not bound to happen everytime it got exited.
But what should I do if I have the pid?

Once you have the PID, if the process is stuck, then send it a signal to cause it to produce a core dump and that will show you what's stopping progress on the forked process. I showed that in my initial comment part d) #33704 (comment) but also here again

kill -s QUIT XXXX

where XXXX is the printed out PID of the process that's stuck.

Maybe I didn't express the problem clearly.
I have two programs one called A,the other called B,I used exec.CombinedOutput to execute B in program A but when B got exited(which we can't see it in the os using ps command),A didn't act as expeced. as a result,A got struck in here.

runtime.gopark
runtime.goparkunlock
runtime.chanrecv
runtime.chanrecv1
os/exec.(*Cmd).Wait
os/exec.(*Cmd).Run
os/exec.(*Cmd).CombinedOutput
main.GuardAthenaService

It's A got struck not B,so I don't see the point in kill B which does not exist anymore?

Hi @q2683252 I'm currently on vacation so I can't currently experiment myself unfortunately. Do you how your process (e.g. athena-keeper) is crashing? Does the Go program work if you try and launch something simple like /bin/ls instead?

Can you try Go 1.13rc1? In Go 1.13 we use vfork on s390x (which is what amd64 does) so that might make a difference.

Athena-keeper exited by itself because of someother issues ,it's a serivice which means it will run for a very long time, it wasn't killed by the os or users.I haven't tried using
a simpler program such as ls but when I manually maked athena-keeper exit immediately once it started ,it did not reappear with many times of tr .I think it might relate to the callee process uptime.

I will try go1.13 ,and maybe it takes a long time to reappear.

This is #23019. Closing as dup.