• Work Areas
• Applications
• Data Management
• VO
• Middleware
  • ARC1
  • Operational Guidelines
  • arc.conf (template)
  • Troubleshooting
  • IGTF Trust Anchors
• Security
• Support
• Monitoring
• Management
  • Operational Activities
  • Fair Share Usage

How to Drain your cluster (before scheduling a downtime)

If you are planing a downtime for your cluster you should gracefully drain your cluster before the downtime kicks in. That is, you shouldn't just stop operation of your cluster while (grid-) jobs are still running.

In ARC you can drain your cluster some day(s) before the scheduled downtime by not accepting new jobs. You need to add the following line in your arc.conf: 

[gridftpd/jobs]
...
allownew="no"

 Notice:

After restarting the gridftpd daemon the information system will publish 'no queue' for your cluster. This will also show up in the GridMonitor as an error (which you can ignore).