Differences

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

--- habrok:additional_information:known_issues [2023/04/26 14:12] – [Getting e-mails on job status changes] camarocico
+++ habrok:additional_information:known_issues [2025/01/21 09:27] (current) – camarocico
@@ Line 1: / Line 1: @@
 ====== Known issues ======
-Some functionality, software, useful tools, and other things are still be missing. The things that we are aware of are listed on this page, and we will update the page whenever the status changes or when new issues have been identified. For other kind of issues, please also check the [[http://status.hpc.rug.nl|HPC status page]].
+This page contains a list of known issues, and, if possible, workaround for them. For other kind of issues, please also check the [[http://status.hpc.rug.nl|HPC status page]].
-===== Data is still being synchronized from Peregrine to Hábrók =====
+===== MFA for F-accounts not possible =====
-All directories from Peregrine's ''/home'' and ''/data'' file systems are being copied to ''/mnt/pg-home'' and ''/mnt/pg-data''. This is still ongoing, and we will keep synchronizing these files until Peregrine is shut down. This means that if you still make any changes to your files on Peregrine, they will show up on Habrok as well (with some delay). On Thursday, March 30, we will disable access to Peregrine, and do one final synchronization of the data to Hábrók, which will likely take about a day.
+Unfortunately, setting up MFA for F-accounts turns out to be impossible. Instead, external collaborators should get an OEP (externall staff) identity via the HR department.
-**//Update March 31, 17:00: the migration of the home directories has completed, which means that you should be able to find all your files in ''/mnt/pg-home/$USER''. The migration of ''/data'' is still ongoing.//**
+===== Getting e-mails on job status changes =====
+The Slurm scheduler has not been connected to a mail server and getting mail notifications is not possible.
-**//Update April 1, 21:00: the final sync of Peregrine’s ''/data'' to Hábrók (''/mnt/pg-data'') has completed as well, which means that all files should now be available. //**
+===== Wrong software stack used for interactive jobs =====
+Interactive jobs, submitted as described on [[habrok:advanced_job_management:interactive_jobs|]], may not be using the software stack that was built for the CPU type of the allocated job, which could lead to application warnings or errors. In order to work around this, you can run ''unset SW_STACK_ARCH && module restore'' when the interactive job has started.
-===== Group memberships and access to (Peregrine) group directories =====
+===== Software using Intel MPI fails when using ''mpirun'' =====
-Though we have started making Hábrók accounts, we still need to create all (relevant) groups that were available on Peregrine, and add accounts to the right groups. This means that you may not be able to access any group directories or group software yet, but we will do this in the week of March 27-31. Do note that, even though you may not be able to access it yet, all group data will be migrated from Peregrine to Hábrók and will not be removed.
+This seems to happen for jobs requesting a larger number of nodes, and only with ''mpirun''. Please see this [[habrok:examples:mpi#intel_mpi | Intel MPI]] section for more information about using ''srun'' instead.
-**//Update 2023-04-05//**: Most of the original pg-groups have been recreated on Habrok with the prefix hb-. In case your group is not there, please let us know.
-===== Missing software =====
-We've already installed quite a few (popular) software modules, but not quite as many as were available on Peregrine. We're still installing some new applications (based on the ones that were recently used on Peregrine), usually only the latest available versions. If there is anything in particular that you would like to have installed on short notice, please let us know.
-===== Missing tools that have to be ported =====
-//**Update April 13: a first release of ''jobinfo'' is now available. It's still lacking GPU usage information for GPU jobs.**// \\
-//**Update April 18: the ''hbquota'' tool for showing information about your quotas on different file systems is now available.**//
-===== Web portal not yet available =====
-The web portal (https://portal.hpc.rug.nl) has not been migrated to Hábrók yet.
-===== V100 nodes unavailable =====
-The Peregrine GPU nodes equipped with an Nvidia V100 GPU will be migrated to Hábrók. As they are running in a different infrastructure and different data center, they are not available yet.
-===== GELIFES nodes =====
-The GELIFES nodes have not been migrated to the Hábrók environment yet. Meanwhile, there is some interim capacity, so the ''gelifes'' partition can already be used on Hábrók by GELIFES users.
-===== MFA for F-accounts not possible =====
-At the moment, F-accounts cannot use MFA, and thus cannot log into Hábrók. We are working on a temporary workaround that would enable F-accounts to login to Habrok, and hope to have something available until the end of April. Please keep an eye on this page for further information.
-//**Update: unfortunately, setting up MFA for F-accounts turns out to be impossible. Instead, external collaborators should get an OEP (externall staff) identity via the HR department.**//
-===== Getting e-mails on job status changes =====
-The Slurm scheduler has not been connected to a mail server yet and getting mail notifications is not possible. We are working on this issue, so please keep an eye on this page for updates.
-===== Wrong software stack used for interactive jobs =====
-Interactive jobs, submitted as described on [[habrok:advanced_job_management:interactive_jobs|]], may not be using the software stack that was built for the CPU type of the allocated job, which could lead to application warnings or errors. In order to work around this, you can run ''module update'' when the interactive job has started.