Now I understand the confusion.

The problem Motorola is seeing is actually a race condition: the moment they invoke `c.exec_start(id); pprint.pprint(c.exec_inspect(id)`, the command haven't finished yet and hence dockerd reports `ExitCode: None`.

I played with this a bit and came up with this solution:


import docker

c = docker.Client()
c.create_container(image="rhel7", name="python_test", command="sleep infinity", tty=True, stdin_open=True)
c.start("python_test")
id = c.exec_create("python_test", "ls -lha /", tty=True)
stream = c.exec_start(id, detach=False, stream=True)
for i in stream:
    print(i)  # will print output of the command to stdout
ret = c.exec_inspect(id)
print("Exit code: %s" % ret["ExitCode"])


And running it:


[root@vm ~]# python t.py
total 0
drwxr-xr-x.   1 root root   6 May 18 14:40 .
drwxr-xr-x.   1 root root   6 May 18 14:40 ..
-rwxr-xr-x.   1 root root   0 May 18 14:40 .dockerenv
lrwxrwxrwx.   1 root root   7 Apr  4 17:06 bin -> usr/bin
dr-xr-xr-x.   2 root root   6 Dec 14 17:23 boot
drwxr-xr-x.   5 root root 360 May 18 14:40 dev
drwxr-xr-x.   1 root root  66 May 18 14:40 etc
drwxr-xr-x.   2 root root   6 Apr  4 17:08 home
lrwxrwxrwx.   1 root root   7 Apr  4 17:06 lib -> usr/lib
lrwxrwxrwx.   1 root root   9 Apr  4 17:06 lib64 -> usr/lib64
drwxr-xr-x.   2 root root   6 Dec 14 17:23 media
drwxr-xr-x.   2 root root   6 Dec 14 17:23 mnt
drwxr-xr-x.   2 root root   6 Dec 14 17:23 opt
dr-xr-xr-x. 108 root root   0 May 18 14:40 proc
dr-xr-x---.   1 root root  23 Apr  4 17:12 root
drwxr-xr-x.   1 root root  21 Apr  4 17:12 run
lrwxrwxrwx.   1 root root   8 Apr  4 17:06 sbin -> usr/sbin
drwxr-xr-x.   2 root root   6 Dec 14 17:23 srv
dr-xr-xr-x.  13 root root   0 May 18 14:14 sys
drwxrwxrwt.   7 root root 132 Apr  4 17:08 tmp
drwxr-xr-x.  13 root root 155 Apr  4 17:06 usr
drwxr-xr-x.  18 root root 238 Apr  4 17:06 var

Exit code: 0


When I used the touch command, I was still hitting the race condition, probably because 'touch' doesn't output anything to stdout (is this a bug in dockerd)?


I also recommend reading docker's source code on how it handles exec:

https://github.com/projectatomic/docker/blob/docker-1.13.1/cli/command/container/exec.go
https://github.com/projectatomic/docker/blob/docker-1.13.1/client/container_exec.go


Please let me know whether the proposed solution works for Motorola.

If this is a regression, it is possible that this has changed in docker daemon.

Antonio, would you know more about this?

From my PoV, it's not a problem to add something like this into docker-py's method exec_start:

if not detach:
  while True:
    ret = c.exec_inspect(id)
    if ret["ExitCode"] is not None:
      break
    sleep(0.01)

To me, it really sounds like there is a problem in detach mechanic.

I finally figured this out!

This is a duplicate of https://bugzilla.redhat.com/show_bug.cgi?id=1501782

I was able to reproduce. As soon as I used the update from the bug above, the problem was gone.

Attaching test case for QE:


import docker
import pprint
import time

d = docker.Client()

runs_count = 10

for tty_value in (True, False):
    for i in range(runs_count):
        co = d.create_container(image="rhel7", command="sleep infinity", tty=tty_value)
        d.start(co)
        ex = d.exec_create(co, "touch /test", tty=tty_value)
        d.exec_start(ex, detach=False)
        # sl = float(i) / float(1000)
        sl = 0.001 * (2 ** i)
        time.sleep(sl)
        ins = d.exec_inspect(ex)
        ec = ins["ExitCode"]
        pprint.pprint("[%d] Exit code: %s, tty: %s, sleep: %s" % (i, ec, tty_value, sl))
        if ec is None:
            print "FAILURE"

Tried with:
python-docker-py-1.10.6-4.el7.noarch
docker-1.13.1-68.gitdded712.el7.x86_64

From the #Comment 0

# docker exec -ti python_test ls /
bin   dev  home  lib64	     media  opt   root	sbin  sys   tmp  var
boot  etc  lib	 lost+found  mnt    proc  run	srv   test  usr
# echo $?
0

# docker exec -ti cli_test ls /
bin   dev  home  lib64	     media  opt   root	sbin  sys   tmp  var
boot  etc  lib	 lost+found  mnt    proc  run	srv   test  usr
# echo $?
0

From the #Comment 6
# python 1.py 
'[0] Exit code: 0, tty: True, sleep: 0.001'
'[1] Exit code: 0, tty: True, sleep: 0.002'
'[2] Exit code: 0, tty: True, sleep: 0.004'
'[3] Exit code: 0, tty: True, sleep: 0.008'
'[4] Exit code: 0, tty: True, sleep: 0.016'
'[5] Exit code: 0, tty: True, sleep: 0.032'
'[6] Exit code: 0, tty: True, sleep: 0.064'
'[7] Exit code: 0, tty: True, sleep: 0.128'
'[8] Exit code: 0, tty: True, sleep: 0.256'
'[9] Exit code: 0, tty: True, sleep: 0.512'
'[0] Exit code: 0, tty: False, sleep: 0.001'
'[1] Exit code: 0, tty: False, sleep: 0.002'
'[2] Exit code: 0, tty: False, sleep: 0.004'
'[3] Exit code: 0, tty: False, sleep: 0.008'
'[4] Exit code: 0, tty: False, sleep: 0.016'
'[5] Exit code: 0, tty: False, sleep: 0.032'
'[6] Exit code: 0, tty: False, sleep: 0.064'
'[7] Exit code: 0, tty: False, sleep: 0.128'
'[8] Exit code: 0, tty: False, sleep: 0.256'
'[9] Exit code: 0, tty: False, sleep: 0.512'

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2018:2025