debug queue limits

This policy expands the scope of BC Policy FY05-05 entitled "Queue Names," by introducing standardized minimum limits for wallclock time and number of cores for the Debug Queue.

In order to facilitate an effective debugging environment on HPCMP production systems, the implementation strategy is as follows:

• At least two percent (2%) of the total computation cores available on each HPCMP production machine shall be made available to the Debug Queue.
• A wallclock time limit of at least thirty (30) minutes shall be set for the Debug Queue on all HPCMP production machines.
• The Debug Queue shall allow multiple debug jobs to run simultaneously, provided the resources are available. However, individual users will be allowed a maximum of four simultaneously running debug jobs.

Centers are encouraged to work with users whose debug requirements exceed these debug queue limits (i.e., greater than 2% of available cores and/or greater than 30 minutes of wallclock time limit) to determine a method that both meets the user's needs and minimizes the impact on other users' production jobs.

Users are also encouraged to make use of the Utility Servers, which provide an additional environment for debugging.