VOV: Important: 2016.09u16 sanity check no longer restarts vovresourced

AlanB_22262
AlanB_22262 New Altair Community Member
edited February 2023 in Altair HPCWorks

Help!  Sanity check did not re-start vovresourced!

The 2016.09u16 version of Runtime software changes a long-standing side-effect of the 'sanity check' operation.  It no longer re-starts the resources daemon 'vovresourced'.

We have found that as customers are making more use of job classes, the time taken to run the initJobClass{} section job class definitions makes vovresourced take significant time to start.

So, we want to avoid re-starting it needlessly.

In this release, vovresourced is automatically started once when vovserver starts.

Thereafter, you need to manage it with 'vovdaemonmgr' from the CLI or from the daemons page of the browser UI.

The resources daemon now has a new command-line option  '-initjobclasses' that takes one of three values:

  • normal -- initialize job classes and continue operation
  • only -- only initialize jobclasses, then exit
  • skip -- just start vovresourced, and do not initialize job classes

This 'only' may be used to cause the initJobClass{} sections to be evaluated, e.g to re-compute needed resources, without affecting the running vovresourced.

Remember that you can check the 'pid' value in resourced.pid file for changes to see whether the daemon  has re-started.
 

Tagged:

Categories