Requesting shared /mnt disks in HTCondor jobs

Since 7 September 2026, HTCondor jobs no longer get automatic access to the shared disks under /mnt/. Each job must explicitly request every /mnt/ folder it needs. /home is not affected by this change and remains available to every job as before.

This applies to both plain jobs and jobs running inside a container image.

How to request a folder

Add one line per folder to your submit file:

+<name> = true

<name> is the last part of the folder’s path under /mnt/. For example, to access /mnt/nfs22_auger, add:

+nfs22_auger = true

Example

A job using a container image that needs nfs17 and users_private:

+SingularityImage = "/cvmfs/farm.particle.cz/singularity/fzu_wn-centos8"
+nfs17 = true
+users_private = true
queue

Available folders

Name to request	Path available in the job
nfs17	/mnt/nfs17
nfs17_cache	/mnt/nfs17_cache
nfs19	/mnt/nfs19
nfs22_auger	/mnt/nfs22_auger
nfs25_auger	/mnt/nfs25_auger
nfs26_auger1	/mnt/nfs26_auger1
nfs27_auger1	/mnt/nfs27_auger1
nfs27_auger2	/mnt/nfs27_auger2
nfs27_auger3	/mnt/nfs27_auger3
nfs28_auger1	/mnt/nfs28_auger1
nfs28_auger2	/mnt/nfs28_auger2
nfs28_auger3	/mnt/nfs28_auger3
sst1m_cs1	/mnt/sst1m_cs1
sst1m_cs2	/mnt/sst1m_cs2
users_private	/mnt/users_private

Need a folder that isn’t listed here, or unsure which name matches what you used to use? Contact fzu@rt.cesnet.cz.

What happens if you forget a folder

A job that doesn’t request a folder does not get an error — the folder simply appears empty inside the job. If a job that used to read or write under /mnt/ suddenly finds nothing there, check that the corresponding +<name> = true line is present in the submit file.

concurrency_limits

You no longer need to set concurrency_limits yourself for these folders. It is now computed automatically from the +<name> = true lines in your submit file, and any value you set manually for it is overridden. If you don’t know what concurrency_limits is, you can ignore this section — it doesn’t affect you.

Please request only what you need

Each requested folder counts against a shared usage limit for that storage server, whether or not the job actually reads or writes to it. Requesting folders you don’t need can block other users from that storage.

Checking your current usage

Run my_nfs_mount_jobs.py from ui (no arguments needed) to see a table comparing your own running/idle job counts per folder against everyone else’s combined, plus the concurrent-job limit currently in effect for each folder.

$ my_nfs_mount_jobs.py
              |       |    all users    |   yourname
tag           | limit | running |  idle | running | idle
--------------+-------+---------+-------+---------+-----
total         | 1 500 |     152 | 3 110 |       0 |    0
nfs17         |   100 |       0 |     0 |       0 |    0
nfs19         |   100 |      48 |   103 |       0 |    0
...           |   ... |     ... |   ... |     ... |  ...