Looping solution

Looping Solution

Here is our approach to running jobs at scale on coreHPC.

• 1. on cluster 7 there is a directory with job control language
• 2. there is a cronjob that runs on corehpc.
• 3. once an hour, it iterates over all jobs in the system, and moves them forward.
• 4. the job status of each job is below.
• 5. the supported job control language is below.

job status

• 0 new job request
• 1 job data has started copying to fac
• 2 job data copy to fac has completed successfully
• 12 job data copy to fac failed (and requires attention)
• 3 job has been submitted on coreHPC
• 4 job is running on coreHPC
• 5 job has completed on coreHPC
• 15 job has failed in coreHPC (and requires attention)
• 6 scp of job results to C7 has started
• 7 scp of job results to C7 has completed normally
• 17 scp of job results to C7 has failed (and requires attention)

supported job control language (JCL)

• Each job has a file. The command syntax is as follows

type: docking, chemstep, other
dockfiles: /nfs/exe/work/jji/projects/target1/dockfiles (includes INDOCK)
tranches:  /nfs/exe/work/jji/projects/target1/tranches (one file per jobs, circa 1 TB each)