Differenze

Queste sono le differenze tra la revisione selezionata e la versione attuale della pagina.

--- oph:cluster:messages [2026/03/23 14:28] – [2026-03-30] Electrical maintenance (planned intervention) diego.zuccato@unibo.it
+++ oph:cluster:messages [2026/05/05 07:50] (versione attuale) – Avviso sovratemperatura diego.zuccato@unibo.it
@@ Linea 5: / Linea 5: @@
 <WRAP center round info>
 To report issues, please write **only** to difa.csi@unibo.it including a clear description of the problem ("my job doesn't work" is __not__ a clear description but is what we usually receive...), including jobID, misbehaving node(s), steps to reproduce, etc.
+</WRAP>
+<WRAP center round alert>
+Remember that bld[15-16] are reserved for courses during the day. Jobs launched while not in a lab lesson will be terminated without further notice. If you need to run jobs to prepare an exam, just add:
+  #SBATCH --exclude=bld[15-16]
+to your job script.
 </WRAP>
 ===== 2026 =====
-==== 2026-03-30 ====
+==== 2026-04-05 ====
 <WRAP center round important>
-Possible (hopefully unlikely) service interruption due to removal of electrical bypass installed on 25/12.
+The server room is experiencing overtemperature due to a failed AC: many (not all) nodes are being drained and will be resumed ASAP.
 </WRAP>
+==== 2026-03-30 ====
+Possible (hopefully unlikely) service interruption due to removal of electrical bypass installed on 25/12.
 In case of emergency, the cluster will be shut down without further notice between 08.00 and 09.00 and reopened as soon as possible.
+==== 2026-03-23 ====
+Planned network interruption around 06.00. Network should remain unreachable for about 5 minutes, but the AD integration (the one that lets you authenticate against bastion-nav and ophfe*) could have issues restarting afterwards and will be checked at about 07.45.
+Active connections and file transfers will be dropped. Running jobs should continue working unless they're using remote resources.
 ==== 2026-03-10 ====
 <del>Possible problems due to electrical maintenance. Cluster might poweroff without further warning, even if we've been reassured that won't happen (yeah... sure... hope it's not just like the last time...).