Failed to start docker on an Amazon linux machine - docker

Question

I am using an Amazon linux machine (p2).
I have installed this docker version:
Client:
Version: 17.03.2-ce
API version: 1.27
Go version: go1.7.5
Git commit: 7392c3b/17.03.2-ce
Built: Wed Aug 9 22:45:09 2017
OS/Arch: linux/amd64
I'm not sure, but I think the issue started after killing a screen which ran some docker container
I'm experiencing this error:
sudo docker ps
Gives:
Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
And:
sudo service docker status
Gives:
docker dead but subsys locked
I have tried both:
sudo rm -rf /var/run/docker
sudo rm /var/run/docker.*
I also tried to restart and stop:
sudo service docker start/stop
I also rebooted the EC2 machine

Answer 1

Try this and restart docker
yum update device-mapper-libs
sudo service docker restart

Answer 2

I am also facing the same issue. Although I (sort of) fixed it by issuing sudo service docker stop and sudo service docker start before running anything in docker.
Details: I am using docker in a spot instance, so it is setup everytime I need to perform some task. I create the docker and upload my files without any problem. But when I issue the command to run an uploaded bash script in docker I face this issue of docker not running. So before running the script I just stop and start docker. Weirdly, simply doing sudo service docker start or even sudo service docker restart did not solve my problem. I had to specifically issue both start and stop commands. But I don't have enough data points just yet, it is only working from the last couple of days and I am not in a hurry to test this hypothesis (of issuing both commands and not just one).

Answer 3

I had 10 docker containers running an ec2 instance(t2.large), each instance was running on its own service and the whole services are running in a cluster. I updated the timezone of the ec2 instance machine, this required me to reboot the instance. I rebooted the instance, this problem surfaced. First thing I noticed was that ssh into the machine was slower than before, I later realized docker ps was throwing that error, I magically resolved that later to realize that some of the container instances are running but they are not serving any page docker logs -f CONTAINER_ID let me know nginx didn't start due to privilege issues that some of my files that supposed to be created were not created.
I later realized that my magical solution was a really magical solution(most magical solutions are not solutions), all my 10 containers were trying to start at the same time which required more memory space than space my instance could offer, I later had to delete services and recreate them one by one - allow one container to start before creating another one in the same cluster. That was when I had peace. I hope this help somebody.