Parsing Error for `sacct` Node Shorthand
The sacct
command reports the number of nodes for jobs running on node counts that are multiples of 1024 using shorthand notation, such as 1K for 1024, 2K for 2048, and so on. This format causes the parsing of these values as integers to fail.
For example:
$ cin_seff -j 11883099
Traceback (most recent call last):
File "/leonardo/prod/opt/tools/cintools/1.0/none/bin/cin_seff", line 599, in <module>
main()
File "/leonardo/prod/opt/tools/cintools/1.0/none/bin/cin_seff", line 539, in main
data["AllocNodes"] = int(main_info[11])
ValueError: invalid literal for int() with base 10: '1K'
where the output of sacct
is the following:
$ sacct -j 11883099 --format JobID,User,Account,State,AllocCPUS,REQMEM,TotalCPU,ElapsedRaw,MaxRSS,ExitCode,NodeList,AllocNodes,Elapsed,QOS
JobID User Account State AllocCPUS ReqMem TotalCPU ElapsedRaw MaxRSS ExitCode NodeList AllocNodes Elapsed QOS
------------ --------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- -------- --------------- ---------- ---------- ----------
11883099 nwedi000 deste_330+ RUNNING 32768 492800G 00:00:00 2151 0:0 lrdn[0006-0007+ 1K 00:35:51 qos_speci+
11883099.ba+ deste_330+ RUNNING 32 00:00:00 2151 0:0 lrdn0006 1 00:35:51
11883099.ex+ deste_330+ RUNNING 32768 00:00:00 2151 0:0 lrdn[0006-0007+ 1K 00:35:51
11883099.0 deste_330+ RUNNING 32768 00:00:00 2135 0:0 lrdn[0006-0007+ 1K 00:35:35