[package list]

ParaView

License information

ParaView is distributed under the 3-Clause BSD License. Some packages leveraged by ParaView have different licenses. The exact licensr formulation for Paraview and those packages can be found on the "ParaView License" page of the ParaView web site.

User documentation

Warning

ParaView should be part of the central software stack when the data analytics and visualization partition (LUMI-D) is available. For the time being, this partition is not available and ParaView is only available as a user-installable EasyBuild recipe. Due to the LUMI-D unavailability only software (CPU) rendering is available.

ParaView is an open-source, multi-platform data analysis and visualization application. ParaView users can quickly build visualizations to analyze their data using qualitative and quantitative techniques. The data exploration can be done interactively in 3D or programmatically using ParaView’s batch processing capabilities.

ParaView was developed to analyze extremely large datasets using distributed memory computing resources. It can be run on supercomputers to analyze datasets of petascale size as well as on laptops for smaller data, has become an integral tool in many national laboratories, universities and industry, and has won several awards related to high performance computation.

• ParaView website
• ParaView tutorials
• ParaView documentation

Installation

ParaView is available as a user-installable EasyBuild recipe. This recipe will only compile the server components (pvserver, pvbatch, pvpython, ...).

Compilation of ParaView takes a long time. We present two options:

• using 32 cores, on the login node, the entire compilation process takes more or less 1 hour 10 minutes but will not consume billing units
• using 128 cores, on a compute node is slightly faster, the entire compilation process takes more or less 45 minutes but will use billing units (96 core-hours).

On a login nodeOn the compute node

To compile on the login node, first prepare your environment:

module load LUMI/22.08
module purge
module load partition/C
module load EasyBuild-user

This setup, prepare your environment to compile a ParaView installation optimized for the compute node (partition/C). By default, ParaView will be installed in $HOME/EasyBuild and will only be accessible by you. If you want to do an installation available to all the members of your project, set the EBU_USER_PREFIX environment variable:

export EBU_USER_PREFIX=/project/project_46XXXXXXX/EasyBuild
module load EasyBuild-user

where you have to change the project_46XXXXXXX value to your actual project number.

The last step is to install ParaView with the command:

eb --parallel=32 -r ParaView-5.10.1-cpeGNU-22.08.eb

Below is a script to compile ParaView on a compute node. Copy the content of this script in a file named build-paraview.sh.

build-paraview.sh

#!/bin/bash

# !!! CHANGE ME !!!
project=project_46XXXXXXX

# Un-comment this export to install in your project directory
# By default installation is done in $HOME/EasyBuild
#export EBU_USER_PREFIX=/project/${project}/EasyBuild

module load LUMI/22.08
module purge
module load partition/C
module load EasyBuild-user

# Compute nodes do not have internet access
# As a consequence we first download all the
# sources on the login node
eb --stop=fetch -r ParaView-5.10.1-cpeGNU-22.08.eb

echo "Submitting job"

sbatch <<EOF
#!/bin/bash
#SBATCH --partition=small
#SBATCH --account=${project}
#SBATCH --job-name=paraview-build
#SBATCH --time=01:00:00
#SBATCH --nodes=1
#SBATCH --mem=0
#SBATCH --cpus-per-task=128

module load LUMI/22.08
module purge
module load partition/C
module load EasyBuild-user

eb --parallel \${SLURM_CPUS_PER_TASK} -r \
  ParaView-5.10.1-cpeGNU-22.08.eb

EOF

echo "Done"

Next, make the script executable and execute it.

chmod +x ./build-paraview.sh
./build-paraview.sh

A job will be submitted to a compute node. ParaView will be usable after this job successfully complete.

Note

The instruction presented above load the partition/C module to generate ParaView binaries optimized for the compute node. These binaries should run without any problem on the login nodes. The default partition module on the login nodes is partition/L so, loading ParaView module from a login node is done by loading the following modules:

module load LUMI/22.08
module load partition/C
module load ParaView

while, from the compute node where partition/C is the default this is done by loading

module load LUMI/22.08
module load ParaView

Usage

The intended usage of ParaView on LUMI is in a client-server configuration:

• the server runs on a login or a compute node on LUMI
• a matching client (same version) runs on your local machine
• all data processing and rendering are handled by the server

Download and install the client

The first step is to install the ParaView client with the same version as the server on your local machine. How to install the client is beyond the scope of this documentation. Generally speaking, it boils down to running the installer that you can download here:

• 5.10.1 client for MacOS (M1)
• 5.10.1 client for MacOS (Intel)
• 5.10.1 client for Windows
• 5.10.1 client for Linux

Start the server on LUMI

Occasional failure on the small partition

When submitting a job to the small partition, the MPI initialization may fail with a "Adress already in use" message. Workaround is to exclude the bad node when submitting the job with the --exclude=nid00XXXX option, where nid00XXXX is the node where the failure occurs. You can exclude multiple nodes with commas separated a list: --exclude=nid00XXXX, nid00YYYY.

On a compute nodeOn a login node

One option to start a ParaView server on a compute node is use an interactive session:

module load LUMI/22.08
module load partition/C
module load ParaView

srun --partition=small --account=project_46XXXXXXX \
     --ntasks=8 --mem=16G --time=01:00:00 \
     --pty pvserver

Here we start a server on 8 processes and 16GB of memory. Don't forget to change the value project_46XXXXXXX to your project number.

Another option is to start the server in a batch job. Here is an example job script:

start-pvserver.job

#!/bin/bash
#SBATCH --job-name=paraview-server
#SBATCH --account=project_46XXXXXXX
#SBATCH --partition=small
#SBATCH --ntasks=8 
#SBATCH --mem=16G
#SBATCH --time=01:00:00

module load LUMI/22.08
module load ParaView

srun --unbuffered pvserver

You can start the batch job by submitting it with sbatch

 $ sbatch start-pvserver.job
Submitted batch job 123456

Then in order to gather the information necessary to create the SSH tunnel (see nect section) we look at the content of the output file.

 $ squeue --me
  JOBID PARTITION     NAME     USER ST  TIME  NODES NODELIST(REASON)
 123456     small paraview  lumiusr  R  0:10      1 nid002092
 $ cat slurm-123456.out
Waiting for client...
Connection URL: cs://nid002092:11111
Accepting connection(s): nid002092:11111

The ParaView server (pvserver) can be run on the login node but you will be limited to only one process: running the server on the login node is only suitable for small dataset.

module load LUMI/22.08
module load partition/C
module load ParaView

pvserver --no-mpi

Once the server has started you will see a message looking like this:

Waiting for client...
Connection URL: cs://uan02:11111
Accepting connection(s): uan02:11111

This message is important because it provide the information necessary to create the SSH tunnel as described in the next section.

Setup the SSH tunnel

Once the server is started, we need to create an SSH tunnel from your local machine to the node where the server is running. Here we assume you have an entry for LUMI in your .ssh/config named lumi.

To create the tunnel, you have to run a command on your local machine that looks like:

ssh -N -L <port>:<host>:<port> lumi

where <host> is the node on which the server is running, i.e., uanXX if you started the server on a login node or nid00XXXX if it runs on a compute node. For example, if at server startup the output was

Waiting for client...
Connection URL: cs://nid002211:11111
Accepting connection(s): nid002211:11111

then, the value of <port> is 11111 and the value of <host> is nid002211 so that the command to create the SSH tunnel will be

ssh -N -L 11111:nid002211:11111 lumi

Connect to the server

Potential connection failure

Sometimes connection to the server fails when the server is executed from a login node. See the information at the end of this section for more information and a workaround.

The last step is to connect the client running on your local machine to the server running on LUMI. To this end, select File → Connect.... You should already have a configuration suitable configuration by default, i.e., a configuration to connect to a server at address cs://localhost:11111. If indeed this is the case, and that 11111 is the port you use, select this configuration and click Connect. The instruction that follow are here in case you need to create a new configuration.

First, go to File → Connect... and click Add Server.

Choose a name for the configuration, for example, LUMI, then:

• for the server type choose Client/Server
• for the host localhost
• for the port, use the value provided at server startup.

Next, click Configure. In the next dialog box, choose Manual for the startup type then click Save.

To connect, select the newly created configuration and click Connect.

The client should now be connected to the server and the server should print Client connected. Another way to check if the connection is successful is to look at the connection information visible from the menu Help → About and selecting the Connection Information tab.

Now that the server is connected, all data processing and rendering are handled by the server instead of the local client.

Disconnecting the client from the server is done via File → Disconnect.... The server will exit when you disconnect, i.e., if you submitted a job, it should end. However, we recommended you to check that this is indeed the case to avoid wasting your cpu-hours allocation.

Potential connection failure

When running the server on the login node, the client may fail to connect with the following error message:

Connection failed during handshake. vtkSocketCommunicator::GetVersion()
returns different values on the two connecting processes
(Current value: 100).

Contrary to what this message seems to indicate, it is not a version mismatch problem. The problem is that the host name is not fully qualified domain name (FQDN) of the login node. A workaround is to use a FQDN of the login node instead of uanXX when creating the SSH tunnel. The FQDN can be obtained with the hostname -A command. For example, for a server running on uan01

 $ hostname -A
ln01-nmn nid50823968 nid50823968 uan01.can

then, create the SSH tunnel using

ssh -N -L 11111:nid50823968:11111 lumi

User-installable modules (and EasyConfigs)

Install with the EasyBuild-user module:

eb <easyconfig> -r

To access module help after installation and get reminded for which stacks and partitions the module is installed, use module spider ParaView/<version>.

EasyConfig:

• EasyConfig ParaView-5.10.1-cpeGNU-22.08.eb, will build ParaView/5.10.1-cpeGNU-22.08

Technical documentation

• ParaView web site

  • ParaView Documentation

• ParaView Git repository

EasyBuild

• ParaView support in the EasyBuilders repository

• ParaView support in the CSCS repository

Version 5.10.1 for CPE 22.08

Created from scratch for LUMI. Some remarks about the build configuration:

• Need CMAKE_SHARED_LINKER_FLAGS=-lpthread to prevent build failure of protobuf
• Built with FFmpeg support as it's available in the central software stack. However, the build fails due to API changes in version 5.x of FFmpeg. Created a patch based of this VTK pull request.