uit-no / hpc-doc Goto Github PK

Sorry to be too late for comments on the pull request.

You should include the MIT license in the repo (maybe as LICENSE_software or LICENSE_code or something like this), since the MIT license expects some copyright holder to be cited:

Copyright <YEAR> <COPYRIGHT HOLDER>

Permission is hereby granted, free of charge [...]

If you only link to the openlicenses.org page, you don't specify the copyright holder. I now, the copyright holder is given in the README, but it might be a good idea to include them also in an own license file.

Do you want to add all contributors as copyright holders or something more general like "Sigma2" or "SDF at IT at UiT"?

To show users how they can install and manage
own Python packages.

This is because we are switching from hard links
to a proper Sphinx toctree structure. The upright/toctree
is new, the slated/links is old. Over time there will be more new
and less old.

In Gaussian documentation there are several references to scripts, which I could not find:
g03parallel.run, g03serial.run, water.inp
They used to be available on /global/apps/gaussian/usage
Espen probably wrote them and should know whether they still exist, are current etc. If they are not, the appropriate rst files describing them will be removed.

And explain how to detect it and why it is typically
a bad sign and penalty for the user.

We only have 328 SL230 nodes, not 372. Numbers of cores need to be scaled accordingly.

It's not clear how to use the different partitions/qos.
So e.g. how to use devel.

We need an example job script. Currently the place holder is guides/job-script-example.rst I kept it for now even though there is nothing there, because I think such an example is important.

Need more documentation on rsync for file transfer.

We should add link to research data application for storage alternatives for migrating users.

Rhe link we have is very old, and also there should be some mention of data classification.

In FAQs there is a link to "Metasenter page on X-server for Windows": <../../../metacenter/metacenter-documentation/metacenter_user_guide/x-server-for-windows>
It came from http://hpc.uit.no/docs/guides/... , but the link doesn't work any more. Shall we just remove it?

the $maxpartasks is not defined in the example at:
http://hpc-uit.readthedocs.org/en/latest/faq/faq.html#running-tasks-in-parallel-within-one-job

in the old documentation at docs.notur.no we had some example scripts one could download
http://docs.notur.no/uit/stallo_documentation/user_guide/running-many-short-tasks?searchterm=running+tasks+in

we need to polish this a bit.

hpc-doc/help/writing-support-requests.rst is one of my favorite things, but it's a bit too long to be linkable when someone has an issue and is honestly trying their best. I distilled it down to a 3-point list that isn't so condescending I can use as a link in a reply:
https://scicomp.aalto.fi/triton/help.html#give-enough-information

What do you think?

On the page http://hpc-uit.readthedocs.org/en/latest/news/news.html there is a link to system log http://stallo-adm.uit.no/wordpress/ , which gives a 403 error (permission denied).

Help users with compressing their data in order to save storage space in conjunction of stallo shutdown

there is a lot of csh info at the page:
http://hpc.uit.no/en/latest/guides/linux.html
I suggest that we drop the csh completely and add some lines about bash instead.
The default shell on stallo is bash and can not be changed unless you ask the admin.
There are only a very few dinosaurs that are still using tcsh and they probably know all this already.

It would be nice to have list of environment variables that are visible to a job, maybe with some default value. I mean things like $SLURM_JOBID or $SLURM_ARRAY_TASK_ID. Is there e.g. a variable for the global working directory? I saw that there is a folder made for every job in /global/work/${USER}/ but is the folder name in some variable? Or do I have to create the variable myself?
I guess, there are lists with SLURM variables around on the web, but I thought it would be good to have it all on one spot, maybe including Stallo/Notur/Sigma2-specific variables as well.

Is it possible to have some kind of log of any given run on Stallo (or other Norwegian clusters)? Good would be some kind of summary appended to the SLURM-log (slurm-xxx.out) with the actual time used, maximum memory consumption, maximum CPUs used etc.
Even better would be a more detailed analysis of the run with values for the parameters named above every x seconds.
I wrote a tiny script some time ago for my local computer to monitor and log memory consumption and CPU business. Happy to share it if anyone is interested 😃

On page http://hpc-uit.readthedocs.org/en/latest/general_info.html there is a <./getting_access> place holder, which does not point to anything. To local users we send the link https://uit.no/ansatte/organisasjon/artikkel?p_document_id=299809&p_dimension_id=88223&p_menu=49281 , but it seems also outdated. Can we put up somewhere a new form? Also, is this https://www.metacenter.no/mas/application/project/pl/ a good link for the national applicants? Or is this better: https://www.notur.no/national-access ?

we need to check the status of hpn-ssh on stallo and fix the documentation accordingly.
I suspect hpn-ssh is obsolete and we should remove most of the documentation.

I suggest including the notur example scripts into this overview also. Will clean up that stuff, but the reference to /global/apps/examples should be edited since this folder do not exist.