In the accounting logs the resources_used.walltime does not appear to equal endtime minus the starttime.
Often a 3 second mismatch, can be up to 19 seconds and as low as 2 seconds, but never matches.
Any ideas as to where the lost time is?
Do jobs take that long to launch and exit?
Example Output of a tracejob showing no used walltime:
12/14/2022 17:25:15 L Considering job to run
12/14/2022 17:25:15 S enqueuing into workq, state Q hop 1
12/14/2022 17:25:15 S Job Queued at request of user@hpc, owner = user@hpc, job name = STDIN, queue = workq
12/14/2022 17:25:15 S Job Run at request of Scheduler@hpc on exec_vnode (n229:ncpus=1)
12/14/2022 17:25:15 L Job run
12/14/2022 17:25:33 S Obit received momhop:1 serverhop:1 state:R substate:42
12/14/2022 17:25:34 S Exit_status=-4 resources_used.cpupercent=0 resources_used.cput=00:00:00 resources_used.mem=0b resources_used.ncpus=1 resources_used.vmem=0b resources_used.walltime=00:00:00