ARGO cluster will be shutdown on Monday, March 10, 2014 to perform necessary maintenance.  A number of changes to the cluster configuration will be implemented.  We expect the cluster to be available by end of day, March 13, 2014.

Please do not plan on starting any long computations this week.  Job submission will be disabled at midnight on Saturday March 8, 2014.  Non-administrative access will be disabled Monday, March 10th by 10am and all jobs still running will be killed at that time.

Planned changes:

  • All nodes will be upgraded to the latest OS revision and patches.
  • Bright Cluster Manager will be upgraded to the latest version.
  • Univa Grid Engine will be upgraded to the latest version.
  • Hadoop will be reconfigured to run the namenode and secondary-namenode processes on the head nodes. This will involve re-formatting “HDFS” filesystem.  We will backup the HDFS filesystem before we make this change, however we would advise users to also backup any critical data separately.

There will also be changes to the queues as follows:

  1. “all.q” will have a 4 hour wall clock time limit.
  2. A new subordinate queue called “all-LowPri.q” will be created with unlimited time and a lower priority.  Jobs running in this queue will be suspended if slots are required in other higher priority queues.
  3. “all-HiPri.q” will now be available to all users and will have a 1 hour wall clock time limit.