question Re: What is shared between multiple copies of the Docker Sandbox? in Archives of Support Questions (Read Only)

What is shared between multiple copies of the Docker Sandbox?

myoung — Fri, 07 Oct 2016 21:18:48 GMT

I like to have multiple copies of the Hortonworks Sandbox for HDP. Typically these sandbox copies are used for specific demo scenarios. This has worked well with the previous VirtualBox-based sandbox. However, I'm wondering if there are problems when doing this with the Docker-based sandbox. Is there anything shared between docker containers based on a similar image?

I installed the Docker Sandbox using my HCC article here: https://community.hortonworks.com/content/kbentry/58458/installing-docker-version-of-sandbox-on-mac.html
I was working through some of the sandbox tutorials and everything was working ok with Atlas.
At some point, but Docker container crashed.
When I restarted the sandbox, the Atlas UI would not load. See the problem here: https://community.hortonworks.com/questions/60125/atlas-ui-not-working-on-docker-hdp-25-sandbox-afte.html#answer-60449

To get around the problem before I knew the fix, I figured I would start with a completely new sandbox. So I repeated the installation procedure above, but gave the image a new name. Here is the output of docker images:

$ docker images
REPOSITORY          TAG                 IMAGE ID            CREATED             SIZE
sandbox-atlas       latest              1ade17087a83        34 hours ago        15.34 GB
sandbox             latest              fc813bdc4bdd        2 weeks ago         14.57 GB

There are two images. The sandbox image is the original import. The sandbox-atlas image is the new import.

So I create a new sandbox container using the sandbox-atlas image, which is a clean image.
I start up the container and try to load the Atlas UI and I'm surprised to find the exact same error!
I tried creating several different containers and they all behaved the same.

Here is the output of docker ps -a:

$ docker ps -a
CONTAINER ID        IMAGE               COMMAND               CREATED             STATUS                       PORTS               NAMES
cc74035a2d71        sandbox-atlas       "/usr/sbin/sshd -D"   24 hours ago        Exited (255) 4 seconds ago                       hdp25-atlas-demo
381ff4de9d9c        sandbox             "/usr/sbin/sshd -D"   42 hours ago        Exited (0) 35 hours ago                          atlas-demo

You should notice that each container is using a different image.

Is there anything shared between docker containers based on a similar, but different image?

Re: What is shared between multiple copies of the Docker Sandbox?

apathan — Fri, 07 Oct 2016 21:47:28 GMT

@Michael Young

Regarding your question - "Is there anything shared between docker containers?", I was just reading an article on the same and found that..

Your data doesn’t live in the container, it lives in a named volume that is shared between 1-N containers that you define. You backup the data volume, and forget about the container. Optimallyyour containers are completely stateless and immutable.

Source: https://blog.docker.com/2016/03/containers-are-not-vms/

More indepth details: https://docs.docker.com/engine/userguide/storagedriver/imagesandcontainers/

Not sure, if this answers your question but thought to add it here.

Re: What is shared between multiple copies of the Docker Sandbox?

myoung — Sat, 08 Oct 2016 00:30:09 GMT

@Ayub Pathan

This was very helpful information. I was thinking in the lines of VMs, which is wrong for Docker containers. I assumed that if I copied the base image and created a container from it, then the data would not be shared between the two.

What I found out is the container for the sandbox is mounting the /hadoop directory on the "host", which is actually the Hyperkit linux VM (on Mac). So even though I used different images and containers, they were sharing a common directory on the VM.

I'm going to try working around this by mapping my container's /hadoop directory to my local mac project working directory, not the hyperkit vm, to see if that keeps the data separate.

Thank you!

Re: What is shared between multiple copies of the Docker Sandbox?

apathan — Sat, 08 Oct 2016 07:13:14 GMT

@Michael Young Thanks for sharing your findings, this is helpful.

Re: What is shared between multiple copies of the Docker Sandbox?

myoung — Sat, 08 Oct 2016 08:14:27 GMT

I wasn't able to use the local mac directory for the /hadoop mount. The /etc/init.d/startup_script run in the container attempt to chown the users/groups to things that don't exist on my Mac. So I found an alternative approach that I think is working well.

I wrote an article that addresses the shared directory issue: https://community.hortonworks.com/articles/60584/how-to-manage-multiple-copies-of-the-hdp-docker-sa.html