Differences

This shows you the differences between two versions of the page.

--- hpc:faq [2025/09/01 07:39] – Yann Sagon
+++ hpc:faq [2025/12/08 13:25] (current) – Yann Sagon
@@ Line 38: / Line 38: @@
 =?=== Cost ====
+??? I'm using Open XDMoD to check how many hours my group used this year, but the amount is quite different from what appears on the invoice or in ''ug_slurm_usage_per_user.py''.
+!!! For invoicing, we use a Slurm metric called billing. This metric aggregates CPU hours, memory usage, and GPU usage. Unfortunately, Open XDMoD does not currently support this metric and therefore does not take into account GPU type or memory usage. We have added a warning about this in [[accounting#openxdmod|Open XDMoD documentation]]
+??? My group only used our private partition. I heard from a colleague that it is free of charge, yet I still received an invoice.
+!!! Because GPU models and RAM usage vary significantly, we now apply a uniform resource accounting system by converting GPU hours and memory usage into CPU-hour equivalents.
+This metric is called billing. If you own a private partition, you receive a fixed annual allocation of billing units that can be used on any cluster. More details are available here: [[accounting#resources_available_for_research_group|Resources available for research groups]]
+??? Your documentation states the cost per hour is 0.0157 CHF, but my invoice shows 0.02 CHF.
+!!! The price per hour shown on the invoice is rounded for display purposes only. The actual calculation uses the full precision of the rate.
+??? I'm a PI and want to see usage details per user.
+!!! You can use the script ''ug_slurm_usage_per_user.py'' for this purpose. An example is provided in our documentation: Aggregate usage by all users of a given PI. [[accounting#aggregate_usage_by_all_users_of_a_given_pi|Aggregate usage by all users of a given PI]]
 ??? I have no idea why I received your email about 'HPC billing'.
@@ Line 267: / Line 281: @@
 If you need to store a large amount of data, consider using the "Academic NAS" service, which you can find here: Academic NAS.
+??? How can I request for a shared directory
+!!!A shared directory is a directory with access permissions granted to a specific group for sharing data.
+Since June 2025, all new shared directories must use groups declared in the Active Directory of UNIGE.
+To request a shared directory please fill the [[https://dw.unige.ch/openentry.html?tid=hpc | HPC form]] on DW.
+(For Oustsider please refer to your PI/repondant)
 ??? How can I access to a shared directory?
 !!! To access a **shared directory**, you need to be added to the appropriate group.
-Please send an email to [[hpc@unige.ch]] including relevant information (Uusername, Group, private_partion etc...) with the responsible person for the share or partition in CC. The responsible person **must** approve the modification.
+For shared directories using a group from Active Directory (groups beginning with GG or GL - example: GL_S_SCIENCES_POSY_LET, please contact your CI (Correspondant Informatique) or use the [[https://dw.unige.ch/openentry.html?tid=adaccess|ADaccess form]] on DW.
+For old group (share_XXX, private_XXX), please send an email to [[hpc@unige.ch]] including relevant information (Uusername, Group, private_partion etc...) with the responsible person for the share or partition in CC. The responsible person **must** approve the modification.
 ??? How can I copy data from one cluster to another one?
@@ Line 382: / Line 405: @@
 ??? I want to run several time the same job with different parameters
 !!!In that case you can use the **job arrays** feature of SLURM. Please, have a look at the documentation [[hpc:slurm#job_array|Job array]]
+??? Is the nodes from my partition in use?
+!!! You can't just check with the partition name as the node may be in use by another partition such as shared-cpu. Here is an example to check the usage of your comput nodes:
+<code>
+squeue -w $(sinfo --noheader --partition <to be replaced by partition-name> --format="%n" | nodeset -f)
+</code>