NAME
sge_schedule - format of the Grid Engine schedule file
DESCRIPTION
The schedule file is written by the qmaster when MONITOR=1 is set in the params item of sched_conf(5).
Note that the file tends to grow rapidly, and should be truncated/rotated, e.g. with logrotate(1). Also it should probably be placed on local disk, not an NFS-mounted filesystem, for instance, because of the large number of small writes it typically gets - e.g. make a symbolic link into the qmaster spool area (assuming that is local).
See qsched(1) for a tool to process the data and produce information on reservations.
FORMAT
A line containing
::::::::
indicates the start of a scheduling interval. It is followed by lines showing the resource usage considered for the interval. They contain the following fields, delimited by ':'.
- job id
-
The numeric job id;
- task id
-
The task id (1 for serial jobs);
- reason
-
The for listing the utilization. One of the following, depending on the job's state:
- RUNNING
-
Running before the interval;
- SUSPENDED
-
Suspended before the interval;
- MIGRATING
-
To be preempted/migrated (currently unused);
- STARTING
-
To be started;
- RESERVING
-
Having resources reserved through advance or normal resource reservation.
- start time
-
Start of resource utilization, in seconds after the epoch;
- end time
-
End of resource utilization, as above;
- level
-
The level at which resources are considered. One of:
- Q
-
Queue;
- H
-
Host;
- G
-
Global;
- P
-
Parallel environment.
- object name
-
Name of the relevant queue, host, parallel environment, or "global";
- resource name
-
Complex name in use or being reserved;
- utilization
-
Amount of the named complex.
FILES
$SGE_ROOT/$SGE_$CELL/common/schedule
SEE ALSO
sched_conf(5), qsched(1).