WLCG Throughput Call

Other Institutes

Other Institutes


Our next North American throughput meeting. Topics for today include:

  1. perfSONAR v4.0
    • Status and plans for RC
    • Status and plans for release
  2. SC16 discussion and results
  3. Open vSwitch experiences
  4. Update on Analytics
  5. Round-table 
  6. AOB and next meeting

Please use Vidyo to connect.

Attending:  Rob Gardner, Shawn McKee, Marian Babik, Garhan Attebury, Andy Lake, Horst Severini, Philippe Laurens, David Lesny

0) Agenda review - No changes

1) perfSONAR v4.0
     Andy summarized the status of perfSONAR v4.0.  Original release was beginning of December. Stability has been an issue.  Going for
     RC3 early in January.   Testing showing a much more stable version. Not yet in staging (testbed is orange). Lots of bug fixes.

     Face-to-face for perfSONAR developers in January and final released around ~February.

     Discussion of some errors in RC2 build 18 testing.

[Wed Dec 14 15:25:21 2016] [error] [client] access to /var/www/pscheduler-server/pscheduler-api-server.wsgi failed, reason: SSL connection required
[Wed Dec 14 15:25:21 2016] [error] [client] access to /var/www/pscheduler-server/pscheduler-api-server.wsgi failed, reason: SSL connection required
File "/var/www/pscheduler-server/pscheduler-api-server.wsgi", line 10, in <module>
ImportError: No module named jinja2

  - Status and plans for RC3:    Plan is to test thoroughly once RC3 is ready.  Jorge will help pilot.

  - Status and plans for release.    Push for update campaign.  Will need the standalone mesh-config that Soichi is working.    PuNDIT status:  targeting test RPMS
     around the time of RC3.  Goal is to have v1.0 RPMS available shortly after pS v4.0 is released.

2) SC16 discussion and results

      Shawn reported on Caltech booth and multiple 100G circuits, SC16 external network usage record and OSiRIS demo.

3) Open vSwitch experiences

      Shawn reported on Open vSwitch (2.6.1) on USATLAS resources.  Plan is to convert USATLAS production resources at participating sites to using
      Open vSwitch.  Allows traffic shaping and visibility for OpenFlow.   Needs further testing for bullet-proof reconfiguration.   Working with SL6.7 with
      stock kernel and also with mainline (4.8.11) kernel as well as CentOS 7.3.  Instructions pages not ready yet :(

4) Update on Analytics

      Rob reported on disruptions in the UC analytics platform.  Has about 20 users.  Some queries exhausted memory and took down the system.
      Problems have been easy to resolve quickly.   Sometimes re-indexing is required to keep things operating smoothly.      

5) Round-table 
      Garhan mentioned packet-loss at Nebraska (long ago).  Packet loss has gotten worse at Nebraska.   Has narrowed it down to any path going from 
      Nebraska via ESnet.  Tests using owping now showing loss.  Path to ESnet in Chicago goes via Internet2 (AL2S circuit).  Internet2 MaDDash (from Marian)
      is visible here https://pas.net.internet2.edu/maddash-webui/

6) AOB and next meeting
      Garhan:  We have a testbed box from DYNES (perfsonar.unl.edu) that may need reinstalling.  Shawn requested that we try to get it working by
      updating from the staging repo once Andy gets newest fixes in place. If we still have problems we can pursue re-install. Jorge may be contacting
      Garhan if he has issues with the box.

      Marian posted the Pre-GDB indico page https://indico.cern.ch/event/571501/

      Next meeting will be ~end of January 2017.   Look for email on the list about this.

      Send additions or corrections to the mailing list.  

      Happy Holidays everyone!


There are minutes attached to this event. Show them.
The agenda of this meeting is empty