This page provides a collection of frequently asked questions.

Is my HOME already migrated?

You can check using the command pwd. The location in the new file system is: /storage/homefs/$USER, where $USER is your user name.

What can I do, during migration of my HOME?

You can still read your data in your HOME, but please prevent writing new data. There will be a rsync process running to migrate your and other users HOME accounts to the new location. At the end, the new HOMEs will be activated and you get notified. If your $HOME already points to /storage/homefs/$USER, the migration of your HOME is finished. And you can continue working.

I need to share data with my colleges. What can I do?

Starting from May we officially introduce Workspaces. A group shared space. In the meantime you can:

  • Use the still existing institute shared directories /home/ubelix/<instituteID>/shared
  • Ask for becoming a Workspace Beta user (see Workspaces Beta)

I read: HOME quota will be 1TB, what now?

Previously, HOME quota was mostly 3TB. Now we will shift to group shared spaces, where each research group has 10TB in Workspaces free of charge and personal HOME 1TB.

The HOME quota in the new location will be increased temporarily until the Workspaces are established in production (end of May). After the introduction of Workspaces there will be a transition period of one month, where you/your research group manager can create a Workspace and you can migrate the data. Afterwards, the quota will be fixed to 1TB in HOMEs.

What if my HOME is full?

If you reached your quota, you will get strange warning about not being able to write temporary files etc. You can check your quota using the 1. Decluttering: Check for unnecessary data. This could be:

  • unused application packages, e.g. Python(2) packages in $HOME/.local/lib/python*/site-packages/*
  • temporary computational data, like already post processed output files
  • duplicated data
  1. Pack and archive: The HPC storage is a high performance parallel storage and not meant to be an archive. Data not used in the short to midterm should be packed and moved to an archive storage.

In general, we consider data on our HPC systems as research data. Further we consider research data to be shared sooner or later. And we aim to support and enhance collaborations. Therefore, we introduce group shared spaces, called HPC Workspaces. Ask your research group manager to add you to an existing Workspace or create a new one. There will be no quota increase for HOME directories.

Where can I get a Workspace?

Workspaces are still in Alpha testing phase. We will soon switch to Beta testing with users. If you are interested, get in touch with us using a Service Portal request or drop an email.

A research group manager need to create the Workspace, since there are possibilities for charged extensions.

If you want to join an existing Workspace. Ask the Workspace manager or its deputy to add you.

How much will a Workspace cost?

Workspaces itself are free of charge. Every research group has 10TB disk space free of charge, which can be used in multiple Workspaces. If necessary, additional storage can be purchased per Workspace, where only the actual usage will be charged, see Workspace Management

What if our 10TB free of charge research group quota is full?

Your Research group manager or a registered deputy can apply for an additional quota. Actual used quota will be charged.

Job issues

Why is my job still pending?

The REASON column of the squeue output gives you a hint why your job is not running.

The job is waiting for resources to become available so that the jobs resource request can be fulfilled.

The job is not allowed to run because at least one higher prioritized job is waiting for resources.

The job is waiting for another job to finish first (–dependency=… option).

The job is waiting for a dependency that can never be satisfied. Such a job will remain pending forever. Please cancel such jobs.

The job is not allowed to start because your currently running jobs consume all allowed CPU resources for your user in a specific partition. Wait for jobs to finish.


The job is not allowed to start because you have reached the maximum of allowed running jobs for your user in a specific partition. Wait for jobs to finish.

(ReqNodeNotAvail, UnavailableNodes:…)
Some node required by the job is currently not available. The node may currently be in use, reserved for another job, in an advanced reservation, DOWN, DRAINED, or not responding.Most probably there is an active reservation for all nodes due to an upcoming maintenance downtime (see output of scontrol show reservation) and your job is not able to finish before the start of the downtime. Another reason why you should specify the duration of a job (–time) as accurately as possible. Your job will start after the downtime has finished. You can list all active reservations using scontrol show reservation.

Why can’t I submit further jobs?

sbatch: error: Batch job submission failed: Job violates accounting/QOS policy (job submit limit, user’s size and/or time limits)

… means that you have reached the maximum of allowed jobs to be submitted to a specific partition.

Job in state FAILED although job completed successfully

Slurm captures the return value of the batch script/last command and reports this value as the completion status of the job/job step. Slurm indicates status FAILED if the value captured is non-zero.

The following simplified example illustrates the issue:


#include <unistd.h>
#include <stdio.h>
int main (int argc, char *argv[]) {
  char hostname[128];
  gethostname(hostname, sizeof(hostname));
  printf("%s says: Hello World.\n", hostname);


# Slurm options
#SBATCH --mail-user=foo@bar.unibe.ch
#SBATCH --mail-type=END
#SBATCH --job-name="Simple Hello World"
#SBATCH --time=00:05:00
#SBATCH --nodes=1
# Put your code below this line
bash$ sbatch job.sh
Submitted batch job 104

Although the job finished successfully…


knlnode02.ubelix.unibe.ch says: Hello World.

…Slurm reports job FAILED:

bash$ sacct -j 104
       JobID    JobName  Partition    Account  AllocCPUS      State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
104          Simple He+        all                     1     FAILED     45:0
104.batch         batch                                1     FAILED     45:0

Problem: The exit code of the job is the exit status of batch script (job.sh) which in turn returns the exit status of the last command executed (simple) which in turn returns the return value of the last statement (printf()). Since printf() returns the number of characters printed (45), the exit code of the batch script is non-zero and consequently Slurm reports job FAILED although the job produces the desired output.

Solution: Explicitly return a value:

#include <unistd.h>
#include <stdio.h>
int main (int argc, char *argv[]) {
  char hostname[128];
  int n;
  gethostname(hostname, sizeof(hostname));
  // If successful, the total number of characters written is returned. On failure, a negative number is returned.
  n = printf("%s says: Hello World.\n", hostname);
  if (n < 0)
    return 1;
  return 0;
bash$ sacct -j 105
       JobID    JobName  Partition    Account  AllocCPUS      State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
105          Simple He+        all                     1  COMPLETED      0:0
105.batch         batch                                1  COMPLETED      0:0