ParaViewWeb on EC2
We have built Docker images containing ParaView 5.5 which support ParaViewWeb applications. These images currently come in two flavors: One for NVidia GPUs w/ EGL rendering support, and another with support for OSMesa (including the llvm and swr rendering backends). You can find these images here. Each of the flavors mentioned above exposes Visualizer as a standalone deployment.
Example deployment on EC2
For this setup we have used an AWS EC2 instance with an NVidia GPU, and running Ubuntu 16.04. Below we describe the steps. Note that if your instance doesn’t have an NVidia GPU, you can still try the osmesa image. For that, just skip the instructions having to do with graphics card driver installation and nvidia-docker2 package installation, and use the osmesa run example at the bottom of this document. All the rest of the steps should remain the same.
Machine setup
Update system packages:
sudo apt-get update
sudo apt-get upgrade
sudo apt-get dist-upgrade
sudo reboot
Install graphics card drivers
sudo apt-get install nvidia-384
sudo reboot
Check graphics card driver installation:
nvidia-smi -q | head |
To install Docker, first install their official GPG key:
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
Check it by verifying the last 8 characters:
sudo apt-key fingerprint 0EBFCD88 |
Now set up the “stable” repository:
sudo add-apt-repository \ |
Then update the package index and install Docker community edition:
sudo apt-get update
sudo apt-get install docker-ce
Now we can install the nvidia-docker2 package, using a similar approach to the one just above. First install the nvidia-docker gpg key:
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
Add sources from the nvidia-docker sources list to apt:
curl -s -L https://nvidia.github.io/nvidia-docker/ubuntu16.04/nvidia-docker.list | \ |
Update the package index again, install nvidia-docker2, then reload the docker daemon configuration:
sudo apt-get update
sudo apt-get install nvidia-docker2
sudo pkill -SIGHUP dockerd
Verify that it’s working:
sudo docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi |
Setting up a world-facing WebServer
We installed Apache on the instance to act as the front-end for our webapps. All we needed was the apache2 package and some confguration.
sudo apt-get install apache2
Now we can configure that webserver. First enable some extra modules, saving the restart for last:
sudo a2enmod vhost_alias
sudo a2enmod proxy
sudo a2enmod proxy_http
sudo a2enmod proxy_wstunnel
Create a simple virtual host configuration:
sudo vim /etc/apache2/sites-available/001-pvw-forward.conf
Just paste this content there for a start:
<VirtualHost *:80> |
Now create the simple website directory structure and possibly add a basic landing page:
mkdir -p /home/ubuntu/paraviewweb-docker/testsite/www
mkdir /home/ubuntu/paraviewweb-docker/testsite/logs
vim /home/ubuntu/paraviewweb-docker/testsite/www/index.html
Here you could copy the following example landing page, though it’s not really necessary for validation of the ParaViewWeb Docker image:
<!DOCTYPE html> |
Now disable the default site, and enable the one you created above, and finally restart the webserver:
sudo a2dissite 000-default
sudo a2ensite 001-pvw-forward
sudo service apache2 restart
All that’s left is to run the docker image as follows:
sudo docker run --runtime=nvidia -p 127.0.0.1:8081:80 -v <host-data-directory>:/data -ti kitware/paraviewweb:pvw-visualizer-5.5.0 "ws://<ec2-hostname-or-ip[:port]>" |
You will obviously replace <host-data-directory> with some real directory where your datasets are located, and replace <ec2-hostname-or-ip[:port]> with the actual hostname or IP address (and possibly port) of the instance.
Some other run examples follow. To run the osmesa image, you don’t need the nvidia runtime:
sudo docker run -p 127.0.0.1:8081:80 -v <host-data-directory>:/data -ti kitware/paraviewweb:pvw-visualizer-osmesa-5.5.0 "ws://<ec2-hostname-or-ip[:port]>" |
Additionally, extra arguments can be passed to the pvpython process that will be launched. Note that while the session url shown in the examples above (ws://<ec2-hostname-or-ip[:port]>) is normally optional, it must be provided if you want to pass extra arguments to pvpython. So, for example, you could pick the swr rendering backend while preventing loading of ParaView registry values like this:
sudo docker run -p 127.0.0.1:8081:80 -v <host-data-directory>:/data -ti kitware/paraviewweb:pvw-visualizer-osmesa-5.5.0 "ws://<ec2-hostname-or-ip[:port]>" "-dr" "--mesa-swr" |
The quotes around the extra arguments to pvpython are only necessary if those arguments contain spaces or other characters that will treatly specially by the shell interpreting the docker run command.
Try it out
If you chose to create a simple landing page, just point your browser at the running instance now:
http://<ec2-hostname-or-ip>
Otherwise, you can point straight through to ParaViewWeb Visualizer:
http://<ec2-hostname-or-ip>/visualizer
Help and support
Please contact us if you need help with deployment or custom development.