On May 16: Brief HTC System interruption; CentOS7 as default in HTC

[chtc-users@xxxxxxxxxxx]

Greetings CHTC users,

This message is for users of our high-throughput computing (HTC) system for events occurring on Thursday, May 16:

• Scheduled maintenance in one of our server rooms will cause a brief interruption to certain HTC services for a few hours.

• The default HTC job requirement for version will change to '7', from '6 or 7'.

Brief HTC Service Interruption

On Thursday, May 16, there will be a scheduled maintenance downtime that affects one of our server rooms, interrupting access to several HTC System services for up to a few hours. These services include:

• submit-1.chtc.wisc.edu and submit-4.chtc.wisc.edu

• group-specific submit servers administered by CHTC

• about half of our execute servers (including the high-memory servers)

• Gluster file share and transfer server (relevant also for placing/removing files in SQUID; jobs depending on existing SQUID files will be unaffected)

The above outages have the following implications for jobs submitted to the high throughput system:

• Jobs submitted from submit-1, submit-4, and research group-specific submit servers will be interrupted if they are running when the maintenance window begins on Thursday. These jobs will remain in the queue and be re-run automatically after the downtime.
  

• Jobs from any submit server but running on the affected execute servers (including the high-memory servers) will be interrupted when the maintenance window begins on Thursday. These jobs will remain in the queue and be re-run.

• Gluster-dependent jobs that have not started by the morning of May 13 will not be allowed to start running until after the downtime completes on May 16. (Doing so will help make sure that Gluster-dependent jobs are not interrupted before the 72-hour default runtime limit.)

CentOS7 as default HTC requirement

We have largely completed the gradual upgrade of the HTC System to a new version of Linux (CentOS7 from the older Scientific Linux 6, aka âSL6â). So, beginning on May 16, new jobs submitted on CHTC-administered HTC submit servers will automatically require Linux major version 7, by default. Various CHTC guides referring to specific OS versions will be updated accordingly on May 16.

• We anticipate that nearly all users should be able to continue submitting jobs as usual without any changes. Prior to May 16, HTC jobs submitted from CHTC-managed submit servers will continue to run on major version 6 *OR* 7, unless users have specifically required a particular version in their submit file âRequirementsâ line. If you already use the âRequirementsâ line to specify a specific Linux version, your jobs will continue to run ONLY on that version on/after May 16.
• If you would like to continue running on either version of Linux*, you will want to add the following line to your submit files: Requirements = (OpSysMajorVer == 6) || (OpSysMajorVer == 7)
  *Thereâs no âharmâ in continuing to use the above, even as we fully convert the remaining execute servers from version 6 to version 7.
• However, those with jobs that can only run on Linux 6 should immediately work to get their software working on Linux 7, as CHTC will begin transitioning all remaining execute servers (currently less than 25% of the HTC System) from SL6 to CentOS7. Get in touch if you need help with compiling and transitioning.

As always, please email us (chtc@xxxxxxxxxxx) with any questions, or feel free to stop by office hours.

Thank you,

Your CHTC Team

Lauren Michael -ÂResearch Computing Facilitator,ÂUniversity of Wisconsin - Madison

Center for High Throughput Computing;ÂAdvanced Computing Initiative;ÂWisconsin Institute for Discovery;ÂACI-REF

www.tinyurl.com/LMichaelCalendar, Discovery 2264, (608)316-4430