The Berkeley Research Computing high performance computing clusters (Savio, Vector, Cortex) will be upgraded to a new operating system, Scientific Linux 7 (SL7), on Wedensday, January 10, 2018.
Everyone who plans to use Savio after the upgrade needs to take action, or risk the possibility that your software and scripts will not work as expected after the upgrade is completed. See Preparing for the SL7 Upgrade, below.
Why upgrade to SL7?
Cluster nodes have been running the previous version of Scientific Linux, SL6, since Berkeley Research Computing launched. SL7 was released almost three years ago, and SL6 is becoming obsolete. Cluster users have been requesting software that is included in SL7, but is difficult to offer on SL6, with increasing frequency. To see the list of new features available in SL7, please see the RedHat Enterprise Linux 7 page, as SL7 is built from the same source.
Preparing for the SL7 upgrade
Software compiled to run on SL6 will not run on SL7. Software must be recompiled for SL7 to run on the upgraded operating system. Prior to the migration to SL7 (anticipated in the late December 2017 timeframe), all Savio users need to confirm that the software they run is available in the test SL7 environment. Importantly, all users who installed software themselves, or had someone install for them, will need to recompile that software and test all packages, modules, etc.
In addition, scripts that invoke software on SL7 may not function as they did under SL6. It is therefore necessary to run user scripts in order to assure they still function as expected. There are a number of reasons why scripts that successfully invoked software on SL6 might not work on SL7, including differences in filesystem layouts between SL6 and SL7, changes to the hierarchical layout of Savio’s module farm, etc.
To prepare for the SL7 upgrade:
- If you installed software on Savio or had someone else install software for you, please:
- Log into the SL7 test node at sl7.brc.berkeley.edu.
- Check the SL7 module farm to see whether the software & version(s) you use is already compiled and installed by the cluster’s system administrators. For instructions on how to see module information, see our webpage on software usage.
- For any software & version(s) not already available in the module farm, recompile.
- Go on to step 2.
- We are asking every active Savio user to test and confirm that the software, scripts, and workflow they run currently will work in the new environment, and to work out what needs to change in advance of the upgrade. This will assure that researchers are not unexpectedly prevented from running important and time-sensitive research computation after the SL7 migration is complete. Please:
- Log into the SL7 test node, sl7.brc.berkeley.edu, to try out all software, packages, modules, scripts, etc. that you use on the cluster by submitting jobs typical of your workflow to the savio2_knl or the savio2_1080ti partition or the savio2 partition with this option --reservation=sl7testing. Identify any issues you encounter, and request assistance of the BRC team if you are unable to resolve issues yourself.
- Review the SL7 module farm to familiarize yourself with where the packages & modules you use can be found in the new module hierarchy.
SL7 upgrade and Jupyterhub
If you use Jupyter notebooks, please test them on the SL7 Jupyterhub node in advance of the upgrade. You can connect to the SL7 Jupyterhub node by visiting https://sl7.brc.berkeley.edu in your browser. This test node will become the production node in January, so if you notice anything amiss, let us know.
BRC staff will be contacting each group on Savio (both condo contributors and FCA groups) in the coming months to check in on how preparation for the SL7 migration is progressing. BRC staff are happy to provide support where needed. If you run into any difficulties with preparing your software and/or scripts for SL7, please e-mail us at firstname.lastname@example.org.
Getting help with the SL7 upgrade
BRC staff are preparing a guide to installing software on the cluster that may be useful to consult as you recompile your software. For hands-on assistance, you can drop by our weekly office hours Wednesdays from 1:30-3:00 pm or Thursdays 9:30-11:30 am at the Academic Innovation Studio (117 Dwinelle). You can also e-mail us at email@example.com; please be patient with us if it takes a couple days for a response.
We will be holding a hands-on training and software installation clinic on November 1, from 3:30-6 PM. Refreshments will be provided.