Deploying ArangoDB in Docker ToolBox

[Martin Krc]

Hi,

I am using Docker ToolBox on my Windows (Home) machine. I tried to deploy ArangoDB there today, but did not manage to do it in the way that keeps data between restarts of the docker virtual machine. The problem is that when I connect a folder within the virtual machine to a real folder, ArangoDB will complain about file system upon startup (mentioning NFS). My question: Is there any way how to get around this issue?

Martin

[Gergely Brautigam]

On Tuesday, May 19, 2020 at 9:37:03 PM UTC+2, Martin Krc wrote:

Hi,

I am using Docker ToolBox on my Windows (Home) machine. I tried to deploy ArangoDB there today, but did not manage to do it in the way that keeps data between restarts of the docker virtual machine. The problem is that when I connect a folder within the virtual machine to a real folder, ArangoDB will complain about file system upon startup (mentioning NFS). My question: Is there any way how to get around this issue?

Hello.

I was wondering what this has to do with docker on windows or the fact of what the virtual machine does or doesn't? You should have to bother with that. That should be seamless.

In order to handle persisting data, simply run the docker container with a volume mount like this:

-v /tmp/arangodb:/var/lib/arangodb3

 

You can find more instruction on how to send in things on arangodb's docker hub page here: https://hub.docker.com/_/arangodb

Cheers,

Gergely.

Martin

[Kerry Hormann]

Hi Martin,

The key is to use a "persistent container" to store the data.  There are several ways of doing this (most people use a very small image like BusyBox), but I use the ArangoDB container so that I have access to all the tools if necessary.  Here's what I use to create the persistent data container:

docker create \

  --name arangodb-persist \

  --entrypoint /bin/sh \

  -v /var/log/arangodb3 \

  -v /var/lib/arangodb3 \

  -v /var/lib/arangodb3-apps \

  -v /etc/arangodb3 \

  arangodb/arangodb:3.6.3 true

The key for this is changing the "entrypoint" to /bin/sh and issuing the "true" command at the end.  This uses the shell to evaluate the "true' statement and exits without starting the DB services.  Also, I specify the volumes I want to export (with the -v flag).

Then, you can map/use those exported volumes in the application service like this:

docker run \

  --name arangodb \

  --volumes-from arangodb-persist \

  -e ARANGO_ROOT_PASSWORD=somepassword \

  -e ARANGO_STORAGE_ENGINE=rocksdb \

  -v /etc/arangodb3/arangod.db.conf:/etc/arangodb3/arangod.conf \

  arangodb/arangodb:3.6.3 --configuration /etc/arangodb3/arangod.conf

Here, I'm also specifying the arangod.conf file, which I maintain outside the container.

Now, you can start/stop/destroy the "arangodb" application container without losing any data.  Also, you can upgrade the app without upgrading the data container - the built-in tools will be out-of-date, but you would only use those in a dire emergency anyway.

Persistent containers have a lot of other advantages, just browse the interwebs and you'll find a bunch of great info.

Cheers,

Kerry

[Martin Krc]

Hello and thanks for the detailed explanation.

I just wonder -- is this really supposed to work with Docker Toolbox? I understood that the key problem in my case is that Docker Toolbox for Windows is based on a virtual machine (Oracle VirtualBox). The virtual machine has a capability of connecting a local machine folder, but it seems to be doing via NFS or something (and indeed the local machine and virtual machine even have different IPs). So then when I run Docker in the virtual machine and mount a folder which is in turn mapped to my local folder, ArangoDB won't start and is complaining about the file system type. I am not sure if your answer is supposed to address this problem, is it?

Martin

Dne úterý 19. května 2020 22:03:26 UTC+2 Kerry Hormann napsal(a):

[Kerry Hormann]

OK - I see your question in a different light....

This may be a tangent to the question, but it might be an important distinction:  I am using Docker Desktop, not the legacy Toolbox solution.  From what I understand, the Desktop app is very different and uses Hyper-V to run a Linux guest to run the containers and built-in Kubernetes workers.  Having started with the Toolbox app, it is my opinion that the Desktop app is far superior and I would highly recommend using it, but I realize that you may not be able to for one or more reasons (no Hyper-V, incompatibility with other VMx solution, etc.).

But as for your question, I do not believe that you can map a Windows folder/volume to a Linux container, at least not without a LOT of trouble.  You can map volumes from Windows-to-Windows, but not between different platforms/modes.  The messages about "NFS shares" are likely a product of how Toolbox maps volumes (through NFS services rather than I/O mapping), but it's not possible on Desktop either as it gives me an "invalid mode" error when I try.

This is why I suggested using a persistent container (read this, this, or this for more info).  This way, both the app and it's datastore are on the same platform (Linux vs. Windows) so there should be no translation going on (mapping, not translation).  The persistent container won't be running, but as long as you don't delete it, the data won't go away.  Again, you don't need to use the ArangoDB container for this - just about any Linux container will do as long as you define the volumes.  Also, using a persistent container allows you to run tools like commit and save, allowing you to take "snapshots" of the data and/or export it (i.e. backup to .tar file).

Hopefully this better addresses your question.