Guides: Configure and use compute environments

You don’t need an access token if you intend to create SSH key credentials, but you will need it for the Tower Agent credentials.

• The user can create an access token as described here.
• Keep it safe, create a new one if you lose it and delete lost tokens.
• Use descriptive names.
• Don’t share your token with others.
• After you close the token creation window, you will not be able to view/copy the token any more, but you can update it or delete it.

Tower Agent is software that runs on the HPC and communicates with the Seqera API to perform all tasks needed on the HPC, including launching a pipeline and monitoring its execution. For an admin to create a tower agent credential, follow these steps:

1. Navigate to the workspace you want to add credentials to, then click on the Credentials tab.
2. Click on the Add Credentials button under Credentials to create a shared Agent connection ID for the Tower Agent.
3. A wizard interface will appear, with some scripts in the Usage box and fields to complete.





4. Give your credential a descriptive name, this is working at the infrastructure level so we recommend creating different credentials for different environments.
5. Keep Shared agent disabled. Check the shared agent section of this guide for more details.
6. Before adding the credential you will need to run the agent on the compute infrastructure. To do this:
   1. Keep the Agent interface open.
   2. Log in to infrastructure (the HPC).
   3. In theory, you can run the Agent from anywhere on the HPC. See our [best practices](/user-guide/hpc-recommendations) for some recommendations.
   4. Copy the usage script from the Agent interface to any text editor and edit the access token to provide your own token created above and provide the path to the work directory for the Agent.
   5. The work directory for the Agent (provided as a parameter in the command) must exist before running the Agent.
   6. Run the edited script relevant to your infrastructure (PBS or SLURM) bash.
   7. This script will download the Tower Agent script to the current work directory on the HPC and make it executable.
   8. Then, it runs the Agent by providing the connection id and access token.
   9. You can see the Agent running on the terminal. Keep it running.
   10. Back in the Seqera interface, click Add to save your credential.

The steps above will help to create the credentials and understand the Agent’s parameters and how it runs. In practice, this can be better optimised by having all scripts and tokens in config files and bash scripts. See [best practice recommendations](/user-guide/hpc-recommendations). The procedure is also described in the Seqera documentation at Quick Start.

1. Navigate back to the launchpad page for the workspace you want to add a compute environment to.
2. Select the Compute Environments tab in the top navigation bar
3. Select Add Compute Environment button
4. Give the environment a name
   • Make it as descriptive as you like
   • For HPC you may wish to configure different compute environments for different project codes, so you could include this in the environment name to easily identify it. For example, specify the name of the HPC and the relevant project code
   
   
   
   
   
5. Select a platform
   • This will depend on which infrastructure you are running and what workload managers are available there.
   • Examples: For Setonix - SLURM, for Gadi - PBSpro.
6. Select workdirectory
   • This can remain as $TW_AGENT_WORK or specify your own
7. Leave the launch directory field empty.
8. Provide the launch and compute queue (as available on the HPC).
   • We recommend using a queue with long wall time for the launch queue as this is dedicated to the nextflow main process that should run for the whole time of the pipeline execution.
   • Compute queue will be used for the tasks of the pipeline so we recommend using the default queue on the HPC.
   • The compute queue can be reconfigured during the execution through the pipeline configuration options.
9. Select staging options. Under Pre-run script:
   • Here you can provide all commands to be executed on the HPC before launching the pipeline including module loading.
   • Usually you will need to load Nextflow, singularity, conda, awscli .. if any of them will be used.
10. Specify advanced options
    • Under head job submit options add project specific and queue data. This configuration is for the job that runs the main nextflow process.
    • It is recommended to use 2 CPUs and 8 GB memory for this job.
    • Apply head job submit options to compute jobs to ensure your account details get passed on.
11. Click on Add and the compute environment will be created.