Skip to content

Installation via Docker

Stuart Berg edited this page Dec 11, 2018 · 5 revisions

To enable deployment on the cluster and testing in a controlled setting, it is possible to launch DVID via a docker container. Images of DVID can be retrieved from the organization flyem on Docker Hub. There is also an example dataset that is wrapped into a docker volume.

To view the dataset example, one can use the built-in DVID viewer provided by dvid-console (and packaged with the DVID docker container), NeuTu, or simply visit HTTP endpoints using a web browser.

Note: please follow the installation instruction for docker. If run on Mac or Windows, a virtual machine is needed. This virtual machine should have enough disk space to accommodate the sample docker volumes.

Installing and Running DVID

To retrieve the DVID image:

% docker pull flyem/dvid

To run DVID:

% docker run --volumes-from <OPT:VOLUME> -p <PORT>:8000 -it flyem/dvid

where a docker volume can be specified and the port where DVID will be visible.

By default, running this container will create an empty DVID database. One can specify a docker volume that contains a DVID db at /var/dvid/db. Note: it is possible to specify a volume directly from the host machine, but it is recommended that the database be wrapped into a volume container for better portability. These volumes are light-weight and actually stored directly on host filesystem on linux or on the VM that runs the containers on Mac and Windows. Here's more info on docker volumes, and why they're useful for containerized databases.

On linux, DVID can be accessed at 127.0.0.1:< PORT >. For Mac and Windows, DVID is available at the virtual machine address with the specified port. For convenience, the dvid-console is installed with dvid.

Note about RAM

If you're using Docker for Mac (or Windows), then the Docker Engine does not have access to all of the RAM on the host machine. By default, the Docker Engine is configured to use up to 2 GB of RAM. We recommend setting this to at least 4 GB. On Mac: From the Docker menu at the top of your screen, select "Preferences" and find the "Advanced" tab. Set Memory to something reasonable (e.g. at least 4 GB).

Running DVID with a Large Dataset Example

We created a docker volume for a part of the FIB-SEM medulla dataset. This includes image segmentation (not completely edited) and grayscale data. To retrieve the volume:

% docker pull flyem/dvid-fib25db

To run the volume:

% docker run --name fib25db -d flyem/dvid-fib25db

To run DVID with this example:

% docker run --volumes-from fib25db -p 8000:8000 -it flyem/dvid

If you run into issues with "too many open files", you'll have to raise the maximum # of open files allowed by your OS. Follow this set of tuning instructions to make sure dvid can process enough http connections.

Clone this wiki locally