Go to the U of M home page

Pages

Tuesday, August 9, 2016

LATIS Establishes Servers Maintenance Schedule Beginning Labor Day


LATIS is establishing a more regular and frequent schedule for performing maintenance on
Linux systems. Beginning Monday, September 5th (Labor Day), we will reboot systems as needed on the first Monday of the month between 5:00 a.m. and 6:00 a.m.

A reminder of this monthly window will be posted at z.umn.edu/claresearchcomputing and added to the login message you see when connecting to the servers via SSH. Additional background on the servers and a FAQ is available at the end of this email.

If you have any questions, please contact us at help@umn.edu.

Sincerely,
Carla, Chris, David, Eric, Joel, Judy, and Peter
LATIS Research Computing and Engineering Team and Psychology IT




Background

Historically, LATIS patched servers quarterly by announcing a system maintenance window, usually held on a weekend morning. While this worked well enough, its frequency was not ideal  and required additional weekend work by system engineers.

We’ve now established automated patching and can schedule servers to reboot at a time when the disruption is less likely to affect normal working hours.



FAQ

Which systems will be impacted by this change in practice?

  • Linux terminal server (LTS)
  • Research and compute servers (e.g. compute.cla.umn.edu, BLiSS servers)
  • Unix-based file servers (“pkg” directories, clio, thalia)
  • Database and web servers

What if I’m running a process that overlaps with the maintenance window?

Servers will be rebooted, so any running processes will be terminated. If you have a long- running process that will be interrupted by a maintenance window, please use checkpointing to ensure your job can pick up where it left off.

Will any systems remain active during the maintenance window?

A given server will only be rebooted if software patches have made a reboot necessary. So while not every server will reboot every time, many of them will, and you should expect that all servers will be unavailable.

What about Windows systems?

Windows systems such as wts.umn.edu will keep their existing weekly maintenance window of Friday mornings from midnight to 1:00 a.m.  

Will this be the only maintenance window needed for LATIS systems?

This new window accommodates most maintenance needs, but there will still be downtimes for unplanned outages, maintenance that cannot be performed during the regular window, or maintenance of other systems to which our systems are dependent. Whenever possible, we will send a notification in advance of any work that falls outside the regular maintenance window.

Will I be notified prior to each month’s maintenance window?

This is the only scheduled notification planned for the regular maintenance windows. As noted above, we will update the login message on our servers to include a reminder of the regularly scheduled maintenance.