Project

General

Profile

September-09-2015

Slides: https://indico.fnal.gov/conferenceDisplay.py?confId=10107

Meeting Notes

Present

Bo Jayatilaka - OSG Production Support
Chander Sehgal - OSG Production Support
Margaret Votava - FNAL project sponsor
Parag Mhashilkar - Project Lead
Frank Wurthwein - OSG Director
Hyunwoo Kim - project member
Marco Mambelli - project member
Tony Tiradani - HEPCloud Technical Lead/Project Member
Burt Holzman - previous project lead, current Assit Head/ Facilities Coordinator
Dave Mason - CMS Tier1 lead
Brian Bockelman (OSG/CMS) - OSG Liaison for GlideinWMS
Gabriele Garzoglio - HEPCloud Project Manager.
Jeff Dost - OSG Factory Operations
Ken Bloom - USCMS Software and Computing Head
Joe Boyd - FIFE Support
Tanya Levshina - FIFE Support

Communication

  • At present smaller VO frontend admins in OSG should continue to report and raise issues related to GlideinWMS through OSG factory operations

Project Management

  • Next Stakeholders meeting to be scheduled in early January 2016

Technical

  • Brian was asked about the release schedule for issues that are marked urgent or high priority. These 3 issues are already listed for next production release 3.2.12.
  • OSG Factory operations is interested in the feature that simplifies BOSCO entry configuration in the factory. If there are no issues with the backward incompatibility, this issues will be addressed in the 8.2 series
  • CMS and OSG are interested in changes that simplifies factory configuration for operations. Jeff is working with Parag to address this. It is waiting on Jeff's availability.
  • Glideinwms v3.3 series is mostly driven by the HEPCloud facility needs. CMS needs through HEPCloud are adequately represented here. As per Tony these glideinwms changes driven by the HEPCloud does not affect the operations of the global CMS pool.
  • There was a question on CMS’s timeline for feature requests and glideinwms upgrade path. CMS frontend does not depend on the features released for HEPCloud. Before a glideinwms release is deployed in CMS production, it goes through a validation process that may take upto 2 weeks.
  • Chander wanted to know about pilot ramp up time and how fast we can fill a site in OSG. In past OSG observed ~200 cores ramp up at BNL site. Bo will be able to partially address this question in upcoming weeks when OSG starts the scalability limitations of the OSG frontend. Recent times, CMS has ramped up to 100K slots across multiple sites in 4 hours. Brian thinks that most of the OSG ramp up issues can be addressed through configuration settings.