Rescue PBS advice for errors from ARCHER

[christopheredsall]

Good artists copy; great artists steal.
-- Pablo Picasso (allegedly)

Before the ARCHER service closes and they move on to ARCHER2 which has slurm, it might be good to copy the information from their Troubleshooting Guide as I have used this several times when debugging PBS issues on Isambard.

[heaton-dev]

I think these sections are generally applicable to Isambard, possibly some rewording needed:

BATCH SYSTEM	apsched: claim exceeds reservation's node-count	The number of nodes required for an aprun command within a PBS job script is larger than the number of nodes requested with the -l select option.	Change the number of nodes requested in the -l select option to match the number required for the aprun command. If the aprun option -n alone is used, the number of nodes required is the number of processes divided by 24, rounded up. If other options are used to change the number of processes per node (e.g., -N, -S, -j) then the calculation is not so easy: please see the aprun section in the User Guide and the aprun man page.

BATCH SYSTEM	The command 'qstat -f [job id]' shows the information 'comment = Not Running: Insufficient amount of resource vntype (cray_compute != )'	Encountered when the system is full. There are not enough free nodes for the job to run.	Job will run when enough resources become free.

BATCH SYSTEM	The command 'qstat -f [job id]' shows the information 'comment = Not Running: Insufficient amount of resource ncpus (R: 264 A: 234 T: 118320)' (the values for R, A and T will vary)	Encountered when the system is full. There are not enough free nodes for the job to run.	Job will run when enough resources become free.

The rest are specific to ARCHER's implementation / policies.