FAQs & known issues 

• blizzard: How can I login to the system, change my password and login shell?
• blizzard: How can I check my disk quota?
• blizzard: How can I access my GPFS data from outside DKRZ/ZMAW ?
• blizzard: How can I choose which account to use, if I am subscribed to more than one project ?
• blizzard: Why doesn't my LoadLeveler job start running?
• blizzard: Why does my job wait so long before being executed.
• or: Why is my job being overtaken by other jobs in the queue
• blizzard: How can I run a short MPI job using up to 4 nodes?
• blizzard: How can I see on which nodes my job was running, e.g. to tell this to beratung in case of errors?
• blizzard: How can I get a stack trace if my program crashes?
• HPSS: How can I get information about my project's data in the tape archive?
• HPSS: How can I use the HPSS tape archive without typing my password every time, e.g. in scripts or jobs?
• HPSS: How can I use HPSS archive in scripts/jobs?

blizzard: How can I login to the system, change my password and login shell?

Login to the system via:

ssh <user>@blizzard.dkrz.de

back_to_top

blizzard: How can I check my disk quota?

Check your individual quota on $HOME file sytem /pf :

wquota -u <uid>             (e.g. wquota -u a201234)

and your project quota on project file system $WORK :

wquota -j <prj>             (e.g. wquota -j ab0123)

$SCRATCH usage can be checked here.

back_to_top

blizzard: How can I access my GPFS data from outside DKRZ/ZMAW ?

You can use either sftp ...

sftp <user>@blizzard.dkrz.de

or  gridftp from blizzard ...

globus-url-copy file:///work/<prj>/<file> gsiftp://server.example.com/tmp/

or gridftp from outside:

globus-url-copy file:///tmp/testfile gsiftp://gridftp.dkrz.de/scratch/u/u123456

blizzard: How can I choose which account to use, if I am subscribed to more than one project ?

Just insert the following line into your job script:

#@ account_no = <Project>            (e.g.:  #@ account_no = ab0123)

Your default account is stored in file  $HOME/.acct. If you do not specifiy the account in your jobscript, your computer time consumption is charged to this account.
back_to_top

blizzard: Why doesn't my LoadLeveler job start running?

llq -s <jobid>

gives information about status of jobs (may e.g. help if you want to find out why the scheduler does not start your job).

Hint: Check if tasks_per_node  and  task_affinity  match:

#@ tasks_per_node = 64        requires       #@ task_affinity = cpu(1)

#@ task_affinity = core(1)    comes with     #@ tasks_per_node = 32

back_to_top

blizzard: Why does my job wait so long before being executed.

or: Why is my job being overtaken by other jobs in the queue

First make sure your job can be executed at all.

If your job is a valid job there are several possible reasons for it to be queued for a long time and/or to be overtaken ...

• ... later submitted jobs with a higher priority (usually these have used less of their share then your job).
• ... by jobs with lower priority that are sufficiently small and specified a wall clock limit to allows them to be considered for backfilling

For a more detailed description of scheduling policies on blizzard please refer to blizzard -> scheduling.

blizzard: How can I run a short MPI job using up to 4 nodes?

Use job class 'express' by inserting this line into your job script:

#@ class = express

back_to_top

blizzard: How can I see on which nodes my job was running, e.g. to tell this to beratung in case of errors?

Insert this line into your Job script:

uniq $LOADL_HOSTFILE

back_to_top

blizzard: How can I get a stack trace if my program crashes?

The official way to find the location where your program crashed is to run it in a debugger like dbx or inspect a core file with the debugger. A quick way to get the stack trace without the need for a debugger is to insert the following lines at the beginning of your program:

include 'fexcp.h'
call signal(11, xl__trce)

Then compile your program with

-g

or

-qlinedebug

If you want to know the absolute path of the source files mentioned in the trace then add the option

-qfullpath

to your compiler options.

Now if you run your program and it happens to crash because of a segment violation, then it will generate some output like this:

$ ./tracetest 

  Signal received: SIGSEGV - Segmentation violation

  Traceback:
    Offset 0x00000080 in procedure dostuff, near line 24 in file tracetest.f90
    Offset 0x0000004c in procedure tracetest, near line 12 in file tracetest.f90
    --- End of call chain ---

Real debuggers will allow you to get much more information in case the problem is not easily identified.

back_to_top

HPSS: How can I get information about my project's data in the tape archive?

A   list of all files   is located in  your project's home directory  /hpss/arch/<prj>.

back_to_top

HPSS: How can I use the HPSS tape archive without typing my password every time, e.g. in scripts or jobs?

Create a file called .netrc in your HOME directory, which has only read permissions for yourself:

blizzard1% ls -la $HOME/.netrc
-rw------- ..... /home/dkrz/<uid>/.netrc
blizzard$ cat $HOME/.netrc
machine tape login <uid> password <password>

Be aware that special characters like "/ "or "," should be escaped with a backslash, e.g.

machine tape login u123456 password Null\,nix

back_to_top

HPSS: How can I use HPSS archive in scripts/jobs?

Assumed you've got a ~/.netrc file you can use pft in scripts as follows:

#!/client/bin/ksh
export DATA_DIR=/hpss/arch/<prj>/<mydir>
echo "DATA_DIR=  $DATA_DIR"
export filename=test.txt
#
pftp << ENDE
cd $DATA_DIR
bin
dir
get $filename
quit
ENDE
#

back_to_top