Talk:Restarting belt automatically: Difference between revisions
Jump to navigation
Jump to search
(script name) |
No edit summary |
||
| Line 2: | Line 2: | ||
: Hi Lothar. I tested it only on magnitUDE so far, but should work on the both clusters. --[[User:Vardan|Vardan]] ([[User talk:Vardan|talk]]) 14:21, 3 July 2024 | : Hi Lothar. I tested it only on magnitUDE so far, but should work on the both clusters. --[[User:Vardan|Vardan]] ([[User talk:Vardan|talk]]) 14:21, 3 July 2024 | ||
:: I thought so, because you need such a ''chain job'' only on systems with time limited queues/partitions. Regarding your approach, having to specify the script name explicitly (<code>sbatch submit</code>) seems somewhat error-prone to me (forgetting to change the line after renaming the script). An (admittedly clumsy looking ;-) alternative would be<br><code>scontrol show job $SLURM_JOB_ID | awk -F= '/^ *Command/{print $2}' | xargs sbatch</code><br>If one never changes the job name, a simple <code>sbatch $SLURM_JOB_NAME</code> would do the trick. --[[User:Lothar.brendel|Lothar]] ([[User talk:Lothar.brendel|talk]]) 15:34, 3 July 2024 (CEST) | :: I thought so, because you need such a ''chain job'' only on systems with time limited queues/partitions. Regarding your approach, having to specify the script name explicitly (<code>sbatch submit</code>) seems somewhat error-prone to me (forgetting to change the line after renaming the script). An (admittedly clumsy looking ;-) alternative would be<br><code>scontrol show job $SLURM_JOB_ID | awk -F= '/^ *Command/{print $2}' | xargs sbatch</code><br>If one never changes the job name, a simple <code>sbatch $SLURM_JOB_NAME</code> would do the trick. --[[User:Lothar.brendel|Lothar]] ([[User talk:Lothar.brendel|talk]]) 15:34, 3 July 2024 (CEST) | ||
::: Yes, sure. I'm usually not changing the name of the submit file, but of course your option is much better. | |||
Revision as of 00:25, 4 July 2024
Hi Vardan, that applies to magnitude and amplitude, right? --Lothar (talk) 12:09, 3 July 2024 (CEST)
- Hi Lothar. I tested it only on magnitUDE so far, but should work on the both clusters. --Vardan (talk) 14:21, 3 July 2024
- I thought so, because you need such a chain job only on systems with time limited queues/partitions. Regarding your approach, having to specify the script name explicitly (
sbatch submit) seems somewhat error-prone to me (forgetting to change the line after renaming the script). An (admittedly clumsy looking ;-) alternative would bescontrol show job $SLURM_JOB_ID | awk -F= '/^ *Command/{print $2}' | xargs sbatch
If one never changes the job name, a simplesbatch $SLURM_JOB_NAMEwould do the trick. --Lothar (talk) 15:34, 3 July 2024 (CEST)- Yes, sure. I'm usually not changing the name of the submit file, but of course your option is much better.
- I thought so, because you need such a chain job only on systems with time limited queues/partitions. Regarding your approach, having to specify the script name explicitly (