Applicable to:
- Plesk for Linux
Symptoms
- After Docker extension upgrade to 1.6.3 version all containers are automatically published to 127.0.0.1 IP on host:
# docker ps --format '{{.ID}} {{.Names }} {{.Ports}}'
34b5670a1bbad phpmyadmin 127.0.0.1:8080->80/tcp, :::8080->80/tcp
8b25215d2e1b M2DB 127.0.0.1:33306->3306/tcp, :::33306->3306/tcp -
It is not possible to connect to a Docker container remotely
Cause
Internal issue.
Resolution
Note: the issue will not occur after Docker extension update to latest 1.6.4 version, but this is applicable only for non affected servers, if the Plesk server with Docker extension 1.6.3 has already a described issue, even after upgrade of Docker extension to 1.6.4 version, the steps below still should be applied.
- Log into the server via SSH.
- Using the vi text editor open the /etc/docker/daemon.json file.
- Remove the line below:
CONFIG_TEXT: {"ip":"127.0.0.1"}
- Save changes
- Restart Docker service to apply the changes:
# systemctl restart docker
Note: if container is not running as expected after step №5, restart container manually with the command by replacing ContainerID with the ID of Docker container:
# docker container restart ContainerID
Comments
7 comments
Lots of people using public IP in config even if it's a local connection. Often to a redis container for sessions, their sites logins and carts will be broken but may not trigger any monitoring they have as site will still respond ok. What a mess.
Are you going to revert this with another update? edit: seen sense there...
edit: It was no accident, it's documented in the release notes! did no one think what would happen? (https://docs.plesk.com/release-notes/obsidian/change-log/#docker-1.6.3)
That caused a few hours of confused trouble shooting during the night... can we ensure the option is not re-added on the next update please.
Honestly, in my perspective as a DevOps, you guys have made a real super mess for the customer. It would be nice to know why you lock something so we have to unlock it in the console even though we didn't want to, almost everyone uses Docker so the services are accessible from the outside.
Guys I've lost about 3 hours caused by the update released this night.
Plesk extension for docker v. 1.6.4 claims to revert changes made by v 1.6.3 but it simply does not.
I had to manually edit config file as per Ivan's instructions above.
Thanks Ivan, you saved my day.
IMHO things like this should never happen in a production scenario like this.
It is insane that this could be allowed to happen without warning
How could you push such an update automatically to all world? This is an catastrophic failure of QOS. Im really upset.
Why didnt you made the update so that everyone can choose either the container binds to local or public interface?
We have a bunch of docker containers running, and it took a while to figure out that the problem was not the update of the container but with docker. I'm baffled how this passed by even a minimum of QA.
And to only get a "reverted the changes" in the change log (when looking at the docker extension in Plesk) without explaining that the issue may persist without a link to this article is another major screw-up... please do better in the future.
Please sign in to leave a comment.