The next OpenStack upgrade will take place Thursday August 20st.

This will include some architectural changes that will have some small but lasting effects on how you use the system. Please read Atypical Impacts Impacts below for details.

Typical Impacts

As usual, running VMs will remain running but API access to start, stop or otherwise change instances will be down for the day.

Atypical Impacts

USERNAME CHANGE: usernames will become your CSAIL email address rather than simply your CSAIL username. For example, I currently authenticate as 'jon'; after the upgrade I will use 'jon@csail.mit.edu'. Passwords, permissions and project memberships will remain unchanged.

WEB INTERFACE URL: Now horizon.csail.mit.edu. this address was introduced in the last upgrade cycle. After upgrade the old address nimbus.csail.mit.edu will only redirect to the new address. After the next 6-month upgrade this redirect may go away.

REGION NAME CHANGE: There used to be only one openstack region, called "RegionOne". That region has been renamed to "CSAIL_Stata" in anticipation of adding other cloud regions. You may need to update scripts that you have for creating resources that specify a region.

Why So Many Changes

OpenStack continues to be a rapidly evolving software platform with a relatively short support life and upgrades only supported sequentially (no skipping versions). So we normally make updates every six months in Summer and over IAP.

This straight upgrade would be relatively low impact this cycle, but we are making a number of architectural changes to facilitate expansion and collaboration. In early Fall we will be opening a new OpenStack region at the Massachusetts Green High Performance Computing Center (MGHPCC) in Holyoke. This will double (more or less) our current cloud capacity. This expansion is based on donated equipment and its location was initially a matter of space availability.

During early planning for this expansion we entered into a partnership with Mass Open Cloud, who are also operating an OpenStack cloud in the MGHPCC. More on the longer term goals of this collaboration in a separate email, but in the short term this collaboration immediately provides the storage we need to make the expansion donation work, additional compute resources, and additional developer and operations staffing (mostly we are bringing operations expertise and they are bringing developers). Since cloud operation scale well even with doubling our size by adding more staff resources through this collaboration we should be able to dedicate more staff time to bringing new higher level cloud services on line (like DBaaS and Big Data analytics like Spark and Hadoop).

There are some complexity costs to this. Since we will now be operating a multi institution cloud (initially with BU users but potentially Harvard, Northeastern, and UMass in future) we need a way to keep user names for clashing hence the move to email addresses.

-- JonProulx - 10 Aug 2015
Topic revision: 14 Sep 2015, StephenJahl

MIT Computer Science and Artificial Intelligence Laboratory


  • About CSAIL
  • Research
  • News + Events
  • Resources
  • People

This site is powered by Foswiki MIT: Massachusetts Institute of Technology