The ECaccess software includes the service of launching user jobs according to the dissemination schedule (Dissemination schedule) of ECMWF's real-time data and products. This service is also known as TC-1 service, or TC-1 jobs. For more information on TC-1, see Simple time-critical jobs.
| Warning | ||
|---|---|---|
| ||
The HPC (cca//ccb) service in Reading terminated on . The Computing services on ecgate will continue until , though with limited support. Please migrate your TC-1 jobs to ATOS as soon as possible. |
Please find hereafter some important notes on changes with the TC-1 service on Atos in Bologna.
| Table of Contents |
|---|
ECaccess Web Toolkit at ECMWF
To access the ECaccess Web Toolkit on ATosAtos, you will have to load the ecccass eacccess module, e.g.:
| No Format |
|---|
$> module load ecaccess $> ecaccess-event-list |
...
In order to ensure a proper functioning of the ECaccess jobs submissions, including the TC-1 service, it is essential to configure properly the SSH Keys under the user HOME on the Atos systems. See HPC2020: How to connect for more details.
You only need to do this once, and you can do it running the following commandYou could use the following commands to issue a new keys pair and configure them:
| No Format |
|---|
$> ssh-keygen -q -t ed25519 -N '' -C "$USER@ecmwf.int" -f /home/$USER/.ssh/id_ed25519 $> cat /home/$USER/.ssh/id_ed25519.pub >> /home/$USER/.ssh/authorized_keyskey-setup |
The private key shall be kept with an empty password. If you have already defined the SSH keys on Atos, e.g. rsa keys, make sure that you added the public key to the authorized_keys file. Without the SSH keys correctly installed, an ECaccess (TC1) job submission will issue an error message like:
...
As the meteogram database has been moved to Bologna, the Metview meteogram service in Reading will no longer work by default. Rather than fixing the TC-1 jobs in Reading, we strongly recommend users to run these jobs to the Atos systems in Bologna. See example metgram batch job under https://confluence.ecmwf.int/display/UDOC/HPC2020%3A+realtime_metgram.sh. Note that you will need to load the new version of the ecmwf-toolbox.
...
Delete your TC-1 jobs in Reading
When your TC1 TC-1 jobs is up and running on the Atos systems in Bologna, please do stop your TC-1 jobs in Reading, by canceling the job in STANDBY mode with 'ecaccess-job-delete', e.g.:
| No Format |
|---|
$> ecaccess-job-list 22379170 ecgate STDBY 0/1 Oct 26 11:49 [bc06h012 (222)] tc1_06bc.cmd 22345657 ecgate DONE 1/1 Oct 25 11:50 [bc06h012 (222)] tc1_06bc.cmd $> ecaccess-job-delete 22379170 |
Mail notifications currently not working.
SLURM email notifications (directive "#SBATCH --mail-type=") in ECaccess batch jobs on Atos, including TC-1 jobs, are not working at this moment.
TC-1 jobs by TC-2 users
TC-2 users (z<MS> UIDs) cannot currently submit TC-1 jobs on the Atos systems. While we resolve this issue, we invite those users to use their personal UID to launch a TC-1 job, e.g. to trigger the TC-2 ecflow suite.
Common errors
The SLURM directive '#SBATCH --workdir=<directory_name>', commonly used on ecgate in Reading is deprecated. On the ATos systems, you should use '#SBATCH --chdir=<directory_name>' instead.
...
Do give your TC-1 jobs sufficient time with the SLURM directive '#QSUB #SBATCH --time=<HH:MM:SS>'.