Overview:
In order to employ backfilling, the scheduler requires jobs to be submitted with a walltime resource. Walltime estimates are almost always longer than the amount of time requires for the job to run. Users will do this overestimating in order to add extra padding so PBS will not kill their job if the job runs too long. Other users will refuse to submit jobs with a walltime resource due the destructive nature of it. Due to overestimated walltimes, the scheduler's calendar is always a poor reflection of reality.
This feature introduces a new resource called soft_walltime. The scheduler will use the soft_walltime resource in place of the walltime resource when calculating the duration of a job.
Interface 1: soft_walltime resource
- Change Control: Stable
- Permissions: Write: Manager Read: Everyone
- Writable only by managers to avoid users from exploiting it.
- Details:
- Using this resource the admin can set an estimate for how long the job will run. This will be used by the scheduler for the job's duration. The job will not be killed if it exceeds the estimate.
- Since Resource_List.soft_walltime can only be set by a manager, it will likely be set via a queuejob hook or resources_default.
- If soft_walltime is requested, it will be rejected with the following error message: qsub: "Cannot set attribute, read only or insufficient permission Resource_List.soft_walltime" or PBSE_ATTRRO (15003) from the API.
- Interface 2: New PBS error message
- Change Control: Stable
- Details:
- An attempt to combine STF jobs with soft_walltime will be rejected with the error message: "soft_walltime is not supported with Shrink to Fit jobs" or PBSE_SOFTWT_STF (15178 - likely to change during development) from the API
In order to employ backfilling, the scheduler requires jobs to be submitted with a walltime resource. Walltime estimates are almost always longer than the amount of time requires for the job to run. Users will do this overestimating in order to add extra padding so PBS will not kill their job if the job runs too long. Other users will refuse to submit jobs with a walltime resource due the destructive nature of it. Due to overestimated walltimes, the scheduler's calendar is always a poor reflection of reality.
...
An enumeration of places where soft_walltime and hard walltime are used:
...