Frequently Asked Questions (FAQ)

Quick reference guide

If you don’t want to browse through pages of text in search for answers to your questions, you can check out our quick reference guide (cheatsheet) to using PDC resources.


I get chdir /afs/ No such file or directory error on Beskow when running

Beskow does not have access to the afs directory on runtime. Therefore you need to move your executables/scripts to your cfs home directory: /cfs/klemming/scratch/<1st letter username>/<username> and execute the codes/scripts from there.

I get “permission denied” in my home directory and permissions are listed as ??????????

You are probably logging in without a forwardable Kerberos ticket. A forwardable ticket is needed to authenticate with AFS where your home directory is stored. Without it you are only allowed to view files that any AFS user can see.

If using kinit, please ensure that the -f option is included in the command:

kinit -f <username>@NADA.KTH.SE

A forwardable ticket will have the F flag in the output of the command klist -f when running it both on your local machine and on the cluster:

    Principal: <username>@NADA.KTH.SE

  Issued                Expires             Flags    Principal
Jun 23 09:39:25 2021  Jun 30 09:39:25 2021  FIA    krbtgt/NADA.KTH.SE@NADA.KTH.SE

If you only see the ticket on your local machine, but not when logged in to the cluster, ensure that the SSH option GSSAPIDelegateCredentials is set to yes. Options can be set either in the config file or on the command line like this:

ssh -o GSSAPIDelegateCredentials=yes <username>@<cluster>

If using Network Identity Manager, please ensure that the “Can be forwarded to other machines” box is ticked.

I get “Disk quota exceeded” in my project directory even though the storage quota is not full

The quota system is different on Beskow/Tegner and Dardel.

On Beskow/Tegner

The error message “Disk quota exceeded” means that some storage quota is full. Most likely, the file you are attempting to write to is inadvertently accounted to you personal quota instead of the project’s quota. Since the personal quota is just 25 GB it is easily filled. This can be confirmed by observing the output of:

lfs quota -hg $USER /cfs/klemming

You can list any mis-accounted files in the project directory with the following command:

find /cfs/klemming/projects/snic/my_proj/my_subdir -not -group $(stat -c %G /cfs/klemming/projects/snic/my_proj)

This problem usually occurs when a directory or file is moved from personal storage, for instance in nobackup, to the project directory. If a file is moved with mv rather than copied with cp then the file is still counted as belonging to the user rather than to the project. The following command can be run to fix the problem and assign directories and files to the project instead:

fixgroup /cfs/klemming/projects/snic/my_proj/my_subdir

In the above command, replace the directory name with the name of the subdirectory where the “Disk quota exceeded” error is occurring.

Please remember to use cp rather than mv when moving files into the project directory in the future.

On Dardel

Check your project quota with:

lfs quota -hp `stat -c "%g" /cfs/klemming/projects/snic/my_project` /cfs/klemming

Check your personal quota with:

lfs quota -hp $UID /cfs/klemming

There should not be any mis-accounting problems mentioned in regards to Beskow/Tegner on the Dardel system.

I get “Disk quota exceeded” in the scratch area on Beskow/Tegner

Please see the preceding entry about project directories. The same problem can occur in the scratch area. You can list any mis-accounted files in your scratch area with the following commmand:

find /cfs/klemming/scratch/u/username -not -group cfs-scr

Unlike project directories, you cannot run the fixgroup command in the scratch area. To fix the problem please contact PDC support, including the output of the above command.

Please remember to use cp rather than mv when moving files into the scratch area in the future.

When I disconnect and reconnect again to a tmux session, I get “permission denied” and/or my job crashes

It is advantageous to use long-life kerberos tickets in tmux sessions so they are still valid after disconnect/reconnect:

# get a 7-days forwardable ticket
$ kinit --forwardable -l 7d <user>@NADA.KTH.SE

To be able to keep your access to the AFS filesystem in tmux after logging out from your original ssh session you need to have started the original tmux like this:

pagsh bash -c "export KRB5CCNAME=$KRB5CCNAME; afslog; tmux"

On Tegner, you will first have to add the heimdal module to get access to the afslog command:

module load heimdal

What this does is basically create a new so-called PAG (Process Authentication Group), which will hold the AFS authentication tokens for the tmux process. This is needed because otherwise the tokens will be destroyed with the PAG for your ssh session when you log out.

How can I share files with PDC support?

By default only you have access to your files. But when in contact with support it can be useful to share log files, Makefiles, source code, core files etc. There are two ways to do this.

  1. Copy the files to your public directory. The public directory is a space in your home directory where everyone has read access. This method is best for small data that is okay to share with all users of the system.

    cp my_file $HOME/Public
    cp -R my_directory $HOME/Public
  2. Give access permission to the files directly. We provide the command support-access to do this automatically (by setting up Access Control Lists). This method works even for large data, and gives access only to PDC staff. It does not work in project directories.

    support-access my_file
    support-access my_directory


I got “kinit: krb5_get_init_creds: No ENC-TS found” error when trying to run kinit

Write an e-mail asking PDC support ( to extend your Kerberos principal. When this has been done you can continue to login again using the same password as you did before. If you do not have a valid time allocation this will not be done.

How do I reset my Kerberos password?

You can reset your Kerberos password (i.e. your password to log in to PDC) using the kpasswd command. Just type the command into a terminal and you will be prompted for your old password, and then asked to type your new password twice.

I got “krb5_cc_new_unique: Credentials cache file permissions incorrect” error when trying to run kinit

Please check the output of klist command as there may be conflict between Kerberos tickets for different realms. Try kdestroy and then kinit again to get Kerberos ticket for NADA.KTH.SE.

For windows users: If the error persists, please use Network Identity Manager to manage Kerberos tickets. Remember to set default identity by right-clicking the Kerberos principal ending with NADA.KTH.SE, and choosing the Set as default menu item.

When using the KTH Ubuntu computers I lose permission to access my documents after getting my Kerberos ticket

Both PDC systems and KTH Ubuntu systems use Kerberos authentication, but are in different realms. Replacing the login session’s credentials with PDC credentials will destroy the access to your AFS home directory, typically causing applications or the entire login session to crash.

A workaround to the issue is to use pdc-* in front of the commands needed to access PDC systems. This includes: kinit, klist, kdestroy, ssh, scp… The new commands would then be (pdc-kinit, pdc-klist, pdc-kdestroy, pdc-ssh, pdc-scp). These commands work in the same way as the original commands, but stores/looks for the Kerberos ticket in a different location.

Kerberos tickets without address

If you get the error message when you connect to PDC

kinit: krb5_get_init_creds: Incorrect net address


Kerberos V5 refuses forwarded credentials because Read forwarded creds failed: Incorrect net address

this is a sign that you are sitting behind NAT. NAT stands for Network Address Translation and is used when you have several computers (for instance at home) who all have addresses on an internal network but are connected through the same router/base station/broadband modem and for someone looking from the outside behave as if they had been assigned the same address from the internet provider.

That the Kerberos-server (i.e. the Kerberos KDC: Key Distribution Center) complains about “Incorrect net address” is because kinit creates an address with your computers internal address - but seen from the KDC it looks like your computer has the address of the NAT that you are behind.

To resolve this, first try asking for addressless tickets from Kerberos, that is use

kinit --forwardable --no-addresses <username>@NADA.KTH.SE

This can be made default for your computer by adding the following to your Kerberos configuration file, krb5.conf:

no-addresses = yes

See more information at Firewalls and kerberos

kinit: krb5_get_init_creds: Incorrect net address

This is most likely caused by a NAT firewall (such as a wideband router used for most home connections).

Remedy: Go to Firewalls and kerberos and try the –no-addresses option to kinit or – with xyz replaced by the IP number of your external NAT interface. This page should give you the address of the external NAT interface in most (but not all) cases.

Kerberos V5: mk_req failed (Server not found in Kerberos database)

This is most often caused by a malfunctioning name server (such as the ones provided by some home consumer ISPs)

Remedy: You will need to add a file krb5.conf which contains a section [domain_realm] with the correct Kerberos realm information and you will need to use an environment variable to tell Heimdal the name of your config file is (if it is not /etc/krb5.conf). Add this content in the krb5.conf file:

[domain_realm] = NADA.KTH.SE = NADA.KTH.SE

kinit: krb5_get_init_creds: unable to reach any KDC in realm NADA.KTH.SE

If you get this error message you are most probably behind a firewall that blocks communication with our Kerberos servers.

Remedy: Go to Firewalls and kerberos.

Cannot find KDC for requested realm while getting initial credentials

Again this likely due to a firewall.

Remedy: Go to Firewalls and kerberos.

Time is out of bounds

If this happens you probably have time synchronization problem:

<username>@NADA.KTH.SE's Password:
kinit: Time is out of bounds (krb_rd_req)

This problem is caused by lack of synchronization between the system you create your Kerberos ticket on and the one you try to login on using that Kerberos ticket. Kerberos demands a maximum of 5 minutes time difference between the system clocks.

Remedy: There are a number of methods to synchronize clocks between machines. The one we recommend is NTP, a protocol for synchronizing clocks over the internet. Information and software for NTP can be found online. If everything looks right, but it does not work anyway, your computer is probably set up for the wrong timezone or the wrong daylight savings time period.

kinit: krb5_get_init_creds: time skew (370) larger than max (300)

This is again caused by the clock on your system being out of sync with the actual time.

Remedy: See information above under Time is out of bounds.

kinit: krb5_get_init_creds: Clock skew too great Unable to negotiate a key exchange method

This is again caused by the clock on your system being out of sync with the actual time.

Remedy: See information above under Time is out of bounds.

kinit/tcp unknown service, using default port 2120

This is not an error message and has no impact on the functionality of Kerberos under normal circumstances. The message informs the user that the kauth/tcp system service is not registered in the client machine as a known service with an assigned port number. The kauth client program therefore selects the default “standard” connection port 2120 when talking to the PDC Kerberos server. This is the wanted behavior.

On most systems the information where the service to port look up table is located is the file /etc/services. Note that other Kerberos client programs (kx, telnet, rsh) may produce similar messages, but may use other port numbers than 2120 as the correct default.

Client’s entry in database has expired

This message indicates that your Kerberos principal has expired. This happens automatically every other year and means that you can not get any Kerberos tickets and therefore you can not login at PDC.

Remedy: Write an e-mail asking PDC support to extend your Kerberos principal. When this has been done you can continue to login again using the same password as you did before.


When I try to login to Dardel I get the error “Connection closed”

The message “Connection closed” means that you are not authenticated to login to Dardel.

The Dardel cluster is not yet generally available. Only selected test users can log in and run jobs at the moment. This applies even if you have been granted an allocation on Dardel. All Dardel allocations currently have a matching Beskow allocation that can be used until your project is migrated. For projects requiring GPUs a Tegner allocation can be granted as well, please contact support. See Information about the migration from Beskow/Tegner to Dardel for more information.

If you are a test user, follow the usual instructions below:

To troubleshoot configuration file problems please try to log in manually…

ssh -vvv -o GSSAPIDelegateCredentials=yes -o GSSAPIKeyExchange=yes \
-o GSSAPIAuthentication=yes <username>@<cluster>

If this works, the original problem was probably caused by errors in the ssh configuration file, otherwise it will write errors you are experiencing.

When obtaining Kerberos credentials I get the error “Preauthentication failed”

This usually means that the wrong passwords was entered. Please double check that you have entered the correct password for your PDC account. If this still fails, please see I forgot my password, how can I get a new one?

I forgot my password, how can I get a new one?

You can request a new password using our online form. The form requires you to submit a copy of a valid ID card or passport. We will send you the new password by SMS or email.

I cannot login but have valid Kerberos

This can either be that you are not allowed to log in since you are not part of an active time allocation at PDC, or there is an error in your configuration files. First of all you should check if you are part of a time allocation on the cluster you are trying to log in to by looking it up on SUPR (if your project is managed through SUPR), or by asking your principal investigator/Course adminstrator if you have been added as a member to the time allocation. If you are a member of an active time allocation there could be some problems with your configuration files. To troubleshoot configuration file problems please try to log in manually…

ssh -vvv -o GSSAPIDelegateCredentials=yes -o GSSAPIKeyExchange=yes \
-o GSSAPIAuthentication=yes <username>@<cluster>

If this works, the original problem was probably caused by errors in the ssh configuration file, otherwise it will write errors you are experiencing.

Bad configuration option: gssapikeyexchange

This error can show upp when you are ssh into a cluster. For example

ssh -vvv -o GSSAPIDelegateCredentials=yes -o GSSAPIKeyExchange=yes \
-o GSSAPIAuthentication=yes <username>@<cluster>

In this case you are using the wrong ssh executable, and you should be using the one that we recommend on our webpages. Follow instructions at How to Login with kerberos

In order to know what ssh you are using on your computer, you can execute

which ssh

Trying to log in to Tegner gives “Network is unreachable”

For various reasons it can happen that the main login node of Tegner crashes or becomes unreachable, with a typical error message like

$ ssh -X <pdc-username>

ssh: connect to host port 22: Network is unreachable

If you experience this, please report it to support via email: Contact Support.

As a temporary workaround solution, you may still be able to access Tegner via the second login node:

$ ssh -X <pdc-username>


Job submit/allocate failed: Invalid partition or qos specification

Usually this depends on user not adding time allocation, or adding it wrongfully. on salloc you should -A <allocationid> and for batch files:

#SBATCH -A <allocationid>

You can check the time allocations you are a member of with

projinfo [options]

[PE_0]:_pmi_applist_setup:App list is not supported for this job!

This error indicates that an MPI application was launched without using srun. See Run interactively and Queueing jobs for how to run commands either interactively using salloc or through the job queue using sbatch.

I want to test PDC resources and check if it suits my needs

If you want to try out PDC resources to make sure it suits your needs, you will need to apply for a PDC account and choose time allocation: pdc-test-xxxx , the last one being the current year. By doing that you will be able to submit your jobs for an specific period of time and check if it fits your needs. If you decide to run using PDC resources you will then need to apply for a time allocation for yourself.

I do not get any e-mails when my job is started/finished

By default SLURM send a an e-mail when your job has started/finished and if you do not have entered any e-mail address in your script, SLURM will automatically send it to the address in .forward so it is important that you see that this file has an updated e-mail.

This files resides in…

/afs/<first letter username>/<username>/Public/.forward

In order to change your e-mail address you can write…

echo "<YOUR NEW E-MAIL>" > /afs/<first letter username>/<username>/Public/.forward


I do not want to recieve any mails regarding flashnews

You can unsubscribe to the mailing list for flashnews at Flashnews for our resources is also available at