hpc-unibe-ch / hpc-unibe-ch.github.io Goto Github PK
View Code? Open in Web Editor NEWGuides, tutorials and documentation about the central HPC resources
Home Page: https://hpc-unibe-ch.github.io/
Guides, tutorials and documentation about the central HPC resources
Home Page: https://hpc-unibe-ch.github.io/
Describe the bug
Currently the search function seems to broken. At least it's not visible anymore.
To Reproduce
Steps to reproduce the behavior:
Expected behavior
Search field visible and useable.
Actual behavior
No search field appears. Searching the page is not possible.
Is your feature request related to a current shortcoming? Please describe.
I had a student using UBELIX without understanding what nodes are and how to use them (aka, running everything on the submit node). There are two mentions in the guide to not doing this:
at the end of the FAQ: "DO NOT run resource-intensive computations directly on the front-end server (submit hosts). This will have a negative impact on the performance of the whole cluster. Instead, submit such computations as a job to the cluster"
But there the wording is unclear, I would change it for "directly on the submit node".
in the quick start guide: "Finally, itβs time for your first job. To do some work on the cluster, you require certain resources (e.g. CPUs and memory) and a description of the computations to be done"
But there it doesn't say that you shouldn't use the submit node even when you can.
Describe the solution you'd like
Change the wording in the FAQ and add a small section/subsection to the quick start guide about what nodes are and what you should and should not do in the submit node.
Describe the bug
When on a device with a small screen, the title area of the navigation in
drawer mode is dark blue instead of red.
To Reproduce
Steps to reproduce the behavior:
Expected behavior
The background color of the title area of the navigation in drawer should be red.
Actual behavior
The background color of the title area of the navigation in drawer should is blue.
Is your feature request related to a current shortcoming? Please describe.
The new diagram has somewhat to many details that might be confusing for beginners not knowing anything yet.
Describe the solution you'd like
Switch back to the old one and adjust the fact that we don't feature submit.unibe.ch anoymore.
Is your feature request related to a current shortcoming? Please describe.
GitHub Actions offer a way to run code on certain events around code management.
One can imagine that as "server-side webhooks".
Describe the solution you'd like
The task to render the documentation should be automated using a Github Action.
Whenever an admin is happy with a new fix/addition in a feature branch, upon merging
this change in a PR into master, the documentation is automatically built.
The whole docu is "user guide", thus we may consider to restructure the pages.
Well established and maybe more intuitive structure could be similar to:
https://support.nesi.org.nz/hc/en-gb
or
https://user.cscs.ch/
Is your feature request related to a current shortcoming? Please describe.
Users ask if we provide LaTeX. We don't.
Describe the solution you'd like
Do not provide LaTeX but help the users with install instruction to empower them to install their own texlive into their home directory.
Describe the bug
To Reproduce
Expected behavior
Actual behavior
Screenshots
Additional context
Is your feature request related to a current shortcoming? Please describe.
As we move to automated publishing of docs using GitHub actions, these are now
rendered into the gh-pages
branch within this repo hpc-docs and served at
https://hpc-unibe-ch.github.io/hpc-docs/
instead.
The publish.sh
previously rendered to the branch
gh-pages
in the repo hpc-unibe-ch.github.io
, which in turn is served at
https://hpc-unibe-ch.github.io/.
How do we want to fix this?
Describe the solutions
In principle there are two equally interesting solutions:
Solution 1: The holistic approach is to move the current repo hpc-docs
to
hpc-unibe-ch.github.io
and everything is as it was before moving to GitHub
actions. Should we want to have a general hpc section and two documentations
for two clusters, we can easily do this by inserting a new first level dividing
the documentation in three sections and add tabs at the top, see
https://michael.rollis.ch/myitjournal for an example of tabs usage (Home,
Courses, Guides, ...). Using this approach all documentation is held within one
repository.
Solution 2: Another approach could be to have a repo for each section, e.g. general stuff
in hpc-unibe-ch.github.io
, one repo for cluster1 and one for cluster2. These
are then available at:
As you see, the URLs are in principle the same, then why at least two potentially
more repos at all? These repos are independent mkdocs "instances" and have
their own mkdocs.yml
conig file. That means, we can have different settings
like the color, blue for cluster1 and yellow for cluster2. When cluster1 is
decomissioned, we simply unpublish the respective repo and archive/delete it.
The topic "search" is ambivalent. Search is based on the repo, that means
searching over several mkdocs instances is not possible. One can then only
search within one domain and not across all hpc-docs. Is this a drawback or an
advantage? Another drawback are links across instances that then have to be
whole URLs like for any other "external" links.
Maybe the drawbacks are too heavy and we should stick to solution 1, shouldn't we?
Let's discuss this and decide.
Most user open a ticket by writing an email to [email protected].
With ServiceNow we perfer to get requests through the Service portal itself.
Steps to carry out to fullfil the task:
The link to the service portal could be:
Dear HPC team,
can you please check the table with QoS description :
I think there should be not "bdw_debug" but "job_bdw_debug", and the same for bdw_short.
Best,
Nadja
Describe the bug
There are several typos on the page. Write out year in the middle and fix propsal in the last admonition.
ToDo:
Is your feature request related to a current shortcoming? Please describe.
The sbatch option --workdir
is no longer availabler. Only the option --chdir
is still usable.
Describe the solution you'd like
Purge --workdir
from the documentation and instead add --chdir
where appropriate.
from our user Docu:
Occasionally we are asked for images like diagrams illustrating the structure of UBELIX or even pictures of machines/storage or the like. Often this is due to the need to describe UBELIX within a research proposal. To support you with this, we provide a short text about UBELIX and some files to download. You can use all of this within your research proposal.
COMING SOON IN 2020
As a user I would like to know which GPU resources are available and how do I request them.
Partially these information are available, but missing information
Especially after the Slurm queue reordering this part need to be re-evaluated.
Is your feature request related to a current shortcoming? Please describe.
Building the local setup for working on the documentation relies on the UBELIX admin
guide, where the procedure is outlined. This is no accessible for UBELIX users and
also makes the admin do another step.
In addition the local environment generated like this may differ between the editors
due to different/old versions installed. Creating the environment should be less
error prone and more reproducible.
Describe the solution you'd like
Create an environment.yml file where all the corner stones of the setup get installed.
Describe how to install conda, the environment and point to mkdocs serve
in the README.md
This environment.yml file is then also leveraged by the GitHub Action proposed in #28.
In the user documentation, add a detailed description where user can find support.
As a user, I am interested in ways to invest into UBELIX and CPU/GPU/storage resources.
The users no longer log in to the cluster via the load balancer (submit.unibe.ch), but pick a specific login node, e.g.
ssh @submit03.unibe.ch
Is your feature request related to a current shortcoming? Please describe.
The page "GPUs" suffers some amount of typographic and other flaws.
Describe the solution you'd like
Improve the page and catch some general typographic convention stuff if applicable.
Is your feature request related to a current shortcoming? Please describe.
As the submit server handling will slightly change, adjustment in the user
guide are needed.
Describe the solution you'd like
Identify places where changes are needed and prepare a new version in a feature branch and merge as soon as the submit servers go productive.
Additional context
See also hpc-unibe-ch/ubelix-controlrepo#248.
Describe how to customize the prompt and add hints to external resources.
Including:
export SQUEUE_FORMAT="%.18i %.9p %.9P %.8j %.8u %.8T %.10M %.9l %.6D %R"
)Link is broken. Correct deep link is:
https://www.scits.unibe.ch/training
the AMD partition will be opened to the public.
It will be most probably restructured in SLURM, but user need to be informed about the existence, the usage and specifications or the AMD nodes and the partition.
Is your feature request related to a current shortcoming? Please describe.
The current default for $PS1 is ok, but there other possibilities
Describe the solution you'd like
Maybe we could provide a primer on $PS1 mangling on a new documentation page that is dedicated to setting up a cozy and sensible user environment that includes more things. The primer could show only the most important as a starting point, illustrate some basic recipes (copy/paste ready) and link to additional resources for the people that want to dive in.
Additional context
There is this id-inside thing that we never wrote an article for. Maybe this paragraph on $PS1 might be interessting for non-HPC users too?
As a user I need more information on how to test if my (long running) application runs efficiently. One way would be Benchmarking.
Extend and improve the Benchmarking documentation.
It should be covered:
There are still information missing:
A declarative, efficient, and flexible JavaScript library for building user interfaces.
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. πππ
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google β€οΈ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.