https://betriebsstatus.tu-dresden.de/en/affected/system_hpc_projectrequest_name/ Incident history github.com/cstate en 2026-05-04T17:00:00+02:00 2026-05-04T17:00:00+02:00 https://betriebsstatus.tu-dresden.de/en/issues/2026-05-04-wartung_downtime_hpc_04-05-26_17_uhr_-_06-05-26_12_uhr/ Mon, 04 May 2026 17:00:00 +0200 https://betriebsstatus.tu-dresden.de/en/issues/2026-05-04-wartung_downtime_hpc_04-05-26_17_uhr_-_06-05-26_12_uhr/ 2026-05-06T12:00:00.000+02:00 <p>DDN and Bull will perform a file system check on /data/horse to identify and repair corrupted files. To facilitate this, all HPC clusters (including login nodes, data movers, and data ports) will be shut down. During this time, no access to data in the HPC file systems will be possible:</p> <pre><code>Downtime Start: May 4, 2026, 17:00 Expected Downtime End: May 6, 2026, 12:00 </code></pre> <p>To make the most of this downtime, Bull, DDN, and ZIH will also carry out additional global maintenance tasks, such as changes on the network and the NFS file systems, and multiple updates. If the work on /data/horse takes significantly longer than the other maintenance tasks, the clusters Capella, Alpha Centauri, and Romeo will be restarted. These clusters have alternative scratch file systems (/data/cat and /data/quokka), allowing jobs to run there. However, only jobs with the appropriate flag for the required file system (-L cat or -L quokka) will be able to start [1]. (Hint: You might as well include the flags for the required file systems in your scripts today.)</p> <p>DDN and Bull will perform a file system check on /data/horse to identify and repair corrupted files. To facilitate this, all HPC clusters (including login nodes, data movers, and data ports) will be shut down. During this time, no access to data in the HPC file systems will be possible:</p> <pre><code>Downtime Start: May 4, 2026, 17:00 Expected Downtime End: May 6, 2026, 12:00 </code></pre> <p>To make the most of this downtime, Bull, DDN, and ZIH will also carry out additional global maintenance tasks, such as changes on the network and the NFS file systems, and multiple updates. If the work on /data/horse takes significantly longer than the other maintenance tasks, the clusters Capella, Alpha Centauri, and Romeo will be restarted. These clusters have alternative scratch file systems (/data/cat and /data/quokka), allowing jobs to run there. However, only jobs with the appropriate flag for the required file system (-L cat or -L quokka) will be able to start [1]. (Hint: You might as well include the flags for the required file systems in your scripts today.)</p> <p>We appreciate your understanding and cooperation as we work to improve the stability and performance of the system.</p> <p>[1] <a href="https://compendium.hpc.tu-dresden.de/data_lifecycle/working/#working-filesystems">https://compendium.hpc.tu-dresden.de/data_lifecycle/working/#working-filesystems</a></p>