Open michaeldayCA opened 6 years ago
Encountered same problem! My docker version is 18.06.1-ce. I added the --init option. Actually, my process tree looks like: |-docker-containe(24408)--init(24439)---python(25306)---{python}(25691) the process with pid 25306 is a zombie process, and it has become the child of init process 24439. But it doesn't get reaped. The 25306 process has released all the host memory, but the cpu and the GPU(this python script trains a DNN) memory are not released. Meanwhile, the 25691 process is a disk-sleep process which is not able to receive any signal as well.... Feel really confused...
I have been experiencing extremely strange behavior lately with docker stack. I can do a docker stack deploy and get a stack up, but when I come back a day later, my docker service ls command shows 2/1 copies of some containers (this is a single node and each container should only have 1 copy). When I run docker stack rm it removes some of the containers but leaves zombies behind that can't be killed and don't show up in docker service ls.
docker kill and docker rm freeze up
Expected behavior
docker stack rm: removes all containers
docker rm -f : kills container
docker kill $(docker ps -q): kills all containers
Actual behavior
docker stack rm: does not remove all containers
docker rm -f : freezes up
docker kill $(docker ps -q): freezes up
Steps to reproduce the behavior
Below is the console output of removing the stack, then finding out that it is not really removed, then seeing that even though docker claims there are no services running, the ps command shows services with the identical stack name as the one just removed, and services with the same names as the ones just removed.
docker stack rm
docker service ls:
docker ps
docker ps
strace docker kill $(docker ps -q):
docker rm -f e9230a01c73c
docker stack rm (2nd try):
Output of
docker version
:Output of
docker info
:Additional environment details (AWS, VirtualBox, physical, etc.)
ESXI connected to vsphere.