ParaViewWeb

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
==============NVSMI LOG==============

Timestamp                           : Thu May 17 19:51:05 2018
Driver Version                      : 384.111

Attached GPUs                       : 1
GPU 00000000:00:03.0
    Product Name                    : GRID K520
    Product Brand                   : Grid

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
pub   4096R/0EBFCD88 2017-02-22
      Key fingerprint = 9DC8 5822 9FC7 DD38 854A  E2D8 8D81 803C 0EBF CD88
uid                  Docker Release (CE deb) <docker@docker.com>
sub   4096R/F273FCD8 2017-02-22

Now set up the “stable” repository:

sudo add-apt-repository \
    "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
    $(lsb_release -cs) \
    stable"

Then update the package index and install Docker community edition:

sudo apt-get update
sudo apt-get install docker-ce

Note that the nvidia-docker2 package is now deprecated (as of Docker 19.03), and the current way of installing nvidia support for gpus within containers is as follows:

# Add the package repositories
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

See the source of that documentation here for full details.

Verify that it’s working:

sudo docker run --rm --gpus all nvidia/cuda:9.0-base nvidia-smi
Wed Nov  6 23:34:43 2019       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.50       Driver Version: 430.50       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:04:00.0  On |                  N/A |
| 30%   46C    P5    29W / 250W |   1864MiB / 11019MiB |     11%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

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>
  ServerName   some.example.com
  ServerAdmin  youremail@example.com
  DocumentRoot /home/ubuntu/paraviewweb-docker/testsite/www
  ErrorLog /home/ubuntu/paraviewweb-docker/testsite/logs/error.log
  CustomLog /home/ubuntu/paraviewweb-docker/testsite/logs/access.log combined

  <Directory /home/ubuntu/paraviewweb-docker/testsite/www>
    Options Indexes FollowSymLinks
    Order allow,deny
    Allow from all
    AllowOverride None
    Require all granted
  </Directory>

  # Handle requests for visualizer
  ProxyPass /visualizer http://localhost:8081/visualizer
  ProxyPassReverse /visualizer http://localhost:8081/visualizer

  # Handle launcher forwarding
  ProxyPass /paraview http://localhost:8081/paraview

  # Handle websocket forwarding
  ProxyPass /proxy ws://localhost:8081/proxy

</VirtualHost>

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>
<html>
  <head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width,initial-scale=1,maximum-scale=1">
    <title>Test Web Site</title>
  </head>
  <body>
    <h1> Test </h1>
    <a href="visualizer">Visualizer</a>
  </body>
</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 --gpus all                     \
    -p 127.0.0.1:8081:80                        \
    -v <host-data-directory>:/data               \
    -e "SERVER_NAME=<ec2-hostname-or-ip[:port]>"  \
    -e "PROTOCOL=ws"                               \
    -ti kitware/paraviewweb:pvw-visualizer-5.5.0

Do not forget to 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 --gpus argument:

sudo docker run                                \
    -p 127.0.0.1:8081:80                        \
    -v <host-data-directory>:/data               \
    -e "SERVER_NAME=<ec2-hostname-or-ip[:port]>"  \
    -e "PROTOCOL=ws"                               \
    -ti kitware/paraviewweb:pvw-visualizer-osmesa-5.5.0

Additionally, extra arguments can be passed to the pvpython process that will be launched by providing the “EXTRA_PVPYTHON_ARGS” environment variable in the docker command. For example, you could pick the swr rendering backend while at the same time preventing loading of ParaView registry values like this:

sudo docker run                                \
    -p 127.0.0.1:8081:80                        \
    -v <host-data-directory>:/data               \
    -e "SERVER_NAME=<ec2-hostname-or-ip[:port]>"  \
    -e "PROTOCOL=ws"                               \
    -e "EXTRA_PVPYTHON_ARGS=-dr,--mesa-swr"         \
    -ti kitware/paraviewweb:pvw-visualizer-osmesa-5.5.0

Note in the above that the extra pvpython args are comma-separated, with no additional spaces inserted.

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.