LWG Minutes 2014-10-22

Agenda

 * SC'14 LWG Meeting
 * Lustre developer meeting
 * Continue discussing plans for Lustre's future

Attendance

 * Chris Morrone (LLNL)
 * Cory Spitz (Cray)
 * Justin Miller (IU)
 * Richard Henwood (Intel)
 * Ruth Klundt (Sandia)
 * James Simmons (ORNL)
 * Colin Faber (Seagate)
 * Paul Sathis (Intel)
 * Patrick Farrell (Seagate)
 * Peter Jones (Intel)
 * Andreas Dilger (Intel)
 * Sebastien Buisson (Bull)

SC'14 LWG Meeting
Reminder that folks need to RSVP for the LWG Meeting. So far only 10 have made it clear they intend to attend. Well under the seating limit of 35, so perhaps no concern there.

Lustre developer meeting
Discussion about feasible dates. First full week of January to be avoided, other weeks an option. Chinese new year possible not until February this year.

Chris needs a rough idea of how many developers can make it in person for planning purposes.

Plans for Lustre's future
Some talk about changing landing scheme (merge topic branches instead of landing individual commits of a topic separately).

Do we have the appetite to replace the work flow we have now? Seems like we have something that works. Maybe we should spend our efforts elsewhere.

Peter: Would moving to OpenSFS hosting the development resources help more people who are not very involved in development get involved in development?

Andreas: I can't claim to be unbiased, but one thing the concerns me is A) duplication of effort, but B) need full time gatekeeper for that branch.

Patrick: Wonder what impediment to landing. Most of development happens at obnoxious landing, but not because it is closed. Just because of complexity and other process issues. Not because it is Intel that hosts it.

James agrees. Not sure that who hosts it that is really the important issue.

Maybe just reduce visibility of intel URLs. For instance, ops manual very visible to users, moving it to opensfs URL might be a lot of bang for the buck.

Justin: can something be done about lustre.org in the mean time? Cory: Make a redirect or pointer to the current web site and Lustre info?

Colin will discuss it with folks at Seagate and see if a link can be made.

Cory: lustre.org is largest impediment to getting a larger development base? lustre.org seems like bigger issue than moving developer web resources to OpenSFS.

Lustre manual on Intel's site is not currently searchable by google because of Intel server current settings.

Would it be easier to start over on the manual? Copying liberally from the existing manual.

Maybe do both in parallel. Create new "start up" manual

asciidoc for the new manual might lower technical barriers to contributions.

Perhaps convert docbook sources to asciidoc, add new introductory

Has OpenSFS considered bringing in a technical writer to improve the manual?

Cory: There are tech writers on staff at Cray and Intel maybe they could be tapped to help out with the lustre manual. Cory will look into what Cray can offer there.

Action Items

 * Provide number of developers likely to attend a developer day (likely in Livermore, CA) to Chris (Everyone)
 * Contact OpenSFS web czar to make automated copy of Lustre manual to OpenSFS hosted server (Chris)
 * Look into availability of Cray technical writer to help with Lustre manual (Cory)
 * See if lustre.org can be changed to present a prominent link to redirect folks to the OpenSFS hosted lustre pages (Colin)