What happens to the data when a docker container crashes

Question

We are trying to move onto Docker for deployment purpose. Our architecture requires to have a redis, a mongodb and several nodejs and java based Docker containers.

So my question is, if suppose the redis/mongodb docker container crashes, do we loose all the data that it had?

We want isolation, but at the same time we don't want to loose data due to malfunction/crashes. Is this even possible to achieve this with Docker or is it something not relevant here?

Any help or comments will be greatly appreciated.

Thanks

Answer 1

The answer is: YES - If a container crashes so that it can not be restored/restarted the data is gone. But, normally containers can be restarted and continued - in that case the data is not lost.

E.g. - the following sequence from the docker docs illustrates how container startup work. Note that the data is not lost here until the container is removed.

# Start a new container
$ JOB=$(sudo docker run -d ubuntu /bin/sh -c "while true; do echo Hello world; sleep 1; done")

# Stop the container
$ sudo docker stop $JOB

# Start the container
$ sudo docker start $JOB

# Restart the container
$ sudo docker restart $JOB

# SIGKILL a container
$ sudo docker kill $JOB

# Remove a container
$ sudo docker stop $JOB # Container must be stopped to remove it
$ sudo docker rm $JOB

Whenever you execute a docker run command you start a new container with fresh data. The data is based on the image you provide and that data is consistent (unless you rebuild the image of course).

So, how should you setup docker to keep your data intact? I think that a good approach is to keep the important data mounted in a volume. Volumes are simply external folders (i.e. a folder from the host system) that holds the data and this data will not be lost even if you reinstall the entire docker daemon.

Example:

docker run -v /some/local/dir:/some/dir/in/redis-container my/redis

This mounts the host folder /some/local/dir as the folder /some/dir/in/redis-container in the running container. If e.g. redis stores its data in that folder you're all set to go and reboots/crashes can be survived.

More info about docker volumes check out the docs. Another great article is the also from the docker website, Managing Data in Containers.

EDIT: After comments I clarified the answer - the data is lost if the container can't be restarted (total crash).