The Workshop Presentation is the recommended starting place to understanding what qlicserver tries to accomplish. It reiterates the basic concept outlined in the original posting to the GridEngine users’ list


The issue at hand is the correct bookkeeping of floating licenses when the GridEngine may or may not share them with other (non-GridEngine) applications. If the float licenses are to be used exclusively on the GridEngine, with no means of external access, you need to read no further. The Administration Guide provides an adequate example of this task. For the balance of the installations (arguably the vast majority), you will need a combination of internal and external license tracking to accomplish the task correctly. This is what the qlicserver code does.

In a Nutshell

Relying on load-sensors to determine license availability simply does not work with the GridEngine notion of comsumable resources.

  1. Instead we query the license server to determine which licenses have been granted and to whom (user and machine) and map these names to equivalent GridEngine resource names. We’ll term these values the ”total” resources available.

  2. Next we query the GridEngine to determine which resources were requested. These resources we can call ”internal” usage.

  3. Any licenses that were granted but that don’t have a correspondence to the internal (GridEngine) usage are deemed ”external” (beyond GridEngine’s control).

  4. The number of licenses we are allowed to manage within GridEngine is simply the difference:

     managed = total - external
  5. Adjust the number of licenses that GridEngine is allowed to manage via the qconf -mattr command.

Remaining Race Condition

This solution eliminates most, but not every race condition:

  1. A delay exists between when a non-GridEngine job starts and its existence is registered via the qconf -mattr procedure.

    Double-checking the license availability within an prolog script can help here. Using the exit code 99 will signal the GridEngine to reschedule the job for the next interval. This extra safety is no longer needed after the next load report interval, at which point the complex_values will have been updated to reflect the non-GridEngine usage.

  2. A race condition can occur when the GridEngine job has started but is slower to occupy the licenses than a non-GridEngine job that starts afterwards.

    There is no general way to preventing this, but some software returns particular exit codes when a license is missing.