Versions Compared

Version Old Version 184 New Version Current
Changes made by

RIS-SVC-conf-publish

RIS-SVC-conf-publish

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • User $HOME directories are intended to allow space for users to make use of the compute cluster, with the knowledge that the Storage Cluster is where data and software will be stored. This is accessed via /storage1, and the Scratch Cluster is accessed via /scratch1.

  • The $HOME directory is required for the Compute Platform(s) to function for users and software often rely on it.

  • Policy dictates that you be limited to 10G of $HOME space.

  • The $HOME directory is NOT backed up and important data should NOT be stored here. Anything you wish to be backed up should be placed in /storage1, this includes scripts.

How do I see what is using up all of my $HOME space?

  • You can use the following command to list out the top 10 (or any number if you replace the 10) files or directories using the most space in your $HOME directory.

  • Make sure the following command is run from your $HOME directory.

Code Block
languagepython
linenumbersfalse
du -hsx .[^.]* * 2>/dev/null | sort -rh | head -10

  • Expected example output.

Code Block
languagepython
linenumbersfalse
800M        .vscode-server
140M        .local
95M work
68M .cache
41M .lsbatch
24M .nv
21M .matlab
20M .npm
20M .config
15M ondemand

Why am I getting a Disk I/O error?

  • This error typically refers to the ability of the job to write a file to a directory.

  • The most common source of the error is a user’s home directory being full.

  • If you encounter this error, please follow the steps below.

How much space is in my Storage Allocation?

  • The Compute Service is connected to the Storage Service via POSIX filesystem mounts.

    • The batch (execution) nodes and condos are connected via cache.

    • The client and interactive nodes are connected directly.

  • The Storage Service provides the SMB interface at smb://storage1.ris.wustl.edu/${STORAGE_ALLOCATION}.

  • The Compute Service provides a POSIX interface via the filesystem path /storage1/fs1/${STORAGE_ALLOCATION}.

  • You can observe available space via SMB mounts with a df command on the mounting workstation, or by the mmlsquota command while logged into the compute system.

...

  • Again, under the Block Limits section, the ‘blocks’ portion is how much you have consumed. The Compute Service uses a caching interface to access the data. Read more about how this affects usage and quota here: cache interfaces

How do I share files in my storage with colleagues?

What’s the best way for me to transfer data?

  • The first method we recommend is to use SMB mounts. You can find more information about connecting at the following link.

  • Our suggested method of transferring data if SMB is not an option is to make use of Globus. You can use Globus in multiple ways. There are links to our Globus documentation below.

How do I request more resources for my job?

  • Requesting more resources for your job means using options that are part of the bsub command. You can find out more information about the bsub options at the following link.

  • Be aware that if the software you use requires special options in order to use these resources, you will need to include those options in your software command as well.

Does RIS offer Docker containers or a repository for them?

  • RIS offers RIS hosted and controlled Docker images. You can find them here.

  • RIS also offers a list of vetted applications where we do not control the Docker image nor host it. You can find that list here.

  • You can request help building a Docker image if you are having trouble via our ticketing system.

  • Software that is used frequently is taken into consideration when creating RIS hosted and controlled Docker images.

  • We currently do not have a public repository for users to host their own images in.

Why can’t I connect to my noVNC image?

  • The first reason this could be happening, is port conflicts.

    • If your job lands on a node that has a job already using the port you are attempting to, you will not be able to connect.

    • You can attempt to launch your job on a new node, or you can change the port you’re using and launch the job again.

  • The second reason this could be happening, is that some department based VPNs are not part of the trusted network that will allow this.
  • If you wish to avoid dealing with ports for GUI based software, you can check out what software we have available through Open on Demand.
  • You can also use port fowarding to get around the second reason for being unable to connect.

Software Debugging Policy

We strive to provide help with software debugging and support to the best of our abilities and time. With that being said, there may be times when we cannot solve an issue related to a specific piece of software or script that is not supported by RIS. In those cases, we will attempt to provide a solution to the problem, but we cannot guarantee that the solution will be successful. We recommend reading this section for more help debugging your software as well as for guidance on software development best practices.