Hot Upgrade

This method reduces upgrade impact -- less downtime for productivity, less obtrusive for the users. The vovserver and vovtaskers are moved separately to the new version.

With this method, vovtaskers with running jobs are temporarily renamed, by appending '_stopped_<timestamp>' to their regular names. This allows active jobs to finish, and restarted vovtaskers to use the regular names.
Note: If you are using a large value of VOV_RELIABLE_TIMEOUT you should suspend job dispatch (see appendix below) for some time, 30-60s, before initiating the cutover.

Follow the steps below for the preparation portion of the procedure:

  1. Download the Accelerator upgrade software.
  2. Install the upgrade software.
  3. Create a separate, temporary test queue (to validate installation).
  4. Validate the installation by starting the temporary test queue with a few vovtaskers on the new version and running test jobs.

Cutover process:

  1. Notify your Accelerator users of the scheduled upgrade.
  2. Get a shell as the Accelerator owner on the Accelerator vovserver host with current-version Altair Accelerator commands in the PATH.
  3. Suspend job dispatching as in appendix below.
  4. Stop the vovserver with the following command:
    ncmgr -q nc-queue stop -freeze
    Note: The vovtaskers with jobs will continue to run and will be renamed. The ones that have no jobs will exit.
  5. Ensure that the shell you are using has a sufficiently high limit for file descriptors.
  6. Source the Altair Accelerator setup file for the new version, or use a separate shell set up for the new version.
  7. Restart the vovserver with the new software version.
    ncmgr -queue nc-queue start
  8. Optional: Restart a subset of the taskers. Use the following command:
    nc cmd vovtaskermgr stop tasker1...taskerN
  9. After the vovtaskers have finished their jobs and exited, run the following command:
    nc cmd vovtaskermgr start tasker1...taskerN