Filter
Top Products
Listing of Tables
at the end of each job by the Partition Manager, pmanager (see Section 4.4).
Table 10.2: Accounting Statistics Table
Field Type Description
name char(16) name of the allocated resource
uid int user ID
project char(16) project name
started UTC time when CPUs were allocated
etime int elapsed time in seconds
atime int CPU seconds allocated
utime int user time in seconds
stime int system time in seconds
cpus int number of CPUs allocated
mem int maximum memory extent in megabytes
pageflts int number of page faults requiring I/O
memint int memory integral in megabyte hours
running 0 | 1 CPUs allocated and running jobs
The etime field stores the elapsed time (in seconds) since CPUs were first allocated to
the resource, including any time during which the resource was suspended.
If a partition is stopped, while a job is running, and the partition is restarted before the
job completes, the etime field will correctly show the total elapsed time of the running
job including the time when the partition was down. If a partition is stopped, while a job
is running, and the job completes before the partition is restarted, the acctstats table
entry will reflect only the time when the partition was running. Any additional time
that the job was running while the partition was down is not included in the acctstats
table entry. If a job is terminated because it exceeds its timelimit or a job is terminated
with rcontrol, the etime field reflects the time for which CPUs were allocated.
The atime field stores the total elapsed time (in seconds) that CPUs have been allocated
– this excludes time during which the resource was suspended, but includes any time
when the partition was down while jobs were running. The value stored is the total for
all CPUs used by the resource.
The utime and stime fields are summed over all processes in the program.
The running field is set to 1 while CPUs are allocated. It is set to 0 when the CPUs are
deallocated, at which point no further updates take place.
The pageflts field shows the number of page faults requiring I/O summed over all of
the processes in the parallel program. It is normally 0. A non-zero and growing value
indicates that the program is paging on some or all nodes.
The RMS Database 10-5