LWG Minutes 2017-06-15

From OpenSFS
Jump to: navigation, search


Cray: Cory Spitz, Ben Evans, Patrick Farrell
ORNL: James Simmons, Dustin Leverman
IU: Ken Rawlings
SuperMicro: Abe Asraoui
HPE: Olaf Weber, Robert Clark
Intel: Andreas Dilger, Joe Gmitter, Peter Jones


New Actions Captured:

  • None

Existing Open Actions:

  • None

Actions Recently Closed:

  • Cory to speak with Chris regarding the open questions on the NRS delay policy docs (LUDOC-366).
  • Cory to get Ben engaged with Henri for review of LU-9021 patches.


2.10 Release Update

  • We are down to 9 issues tracking with FixVersion 2.10, 5 of which have a fix in the ready state and is in gatekeeper testing.
  • One outstanding issue is LU-9073 that James has been working on and Sebastian Buisson providing feedback.
    • The good news is that John fixed the workaround and it is ready. Sebastian has been testing due to having a kerberos setup to use.
  • LU-9049 we have a fix that we believe works and is undergoing further testing.
  • Two issues which are the long poles at the moment are LU-9405 and LU-9485. We are still isolating what the problem is there. These may be become blockers once we understand the issue.
  • On the documentation front, NRS delay policy, Project Quota, and multi-rail have patches in flight that are undergoing refinement.
  • James: We are getting an ARM machine soon and likely will need the last 2 patches from LU-8703. There are two patches left and trying to get reviewers to look at them before code freeze.

Upstream Lustre Client

  • Submitted a bunch of cleanups. James reports that they are thinking of phasing out the 80 character issue.
    • Andreas: The logic for it is really is that it effects the number of terminals you can have open on one screen without line wrapping.
    • James: Linus is thinking about going to 100.


  • We decided to delay server upgrades until after LUG and will look do it before the next LWG call.
  • There is interest in putting together a Lustre tutorial. The thinking on this is to submit it as a BOF session at SC this year. Will start engaging people interested in collaborating on it. Any thoughts?
    • Peter: Is the thinking to have the BOF for new Lustre developers?
      • Ken:Yes
    • Peter: Is anyone on the call attending SC that could participate?
      • Dustin: May be going and might be able to participate.
    • Peter: What do you envision us needing for content? For a BOF it would have to be 60-90 min in length. Is this reasonable?
      • Discussion ensued around thoughts of what is best addressed in getting started. General consensus was it would be difficult to accomplish this in that sort of time frame.
      • If engineers were able to download items ahead of time and prepare, then it comes down to probably about 60 min of actual training time.
      • To be successful at this based on hackathon, a good idea of number of participants is needed and at least enough helpers to have 1 engineer available for every 3 engineers in training.

Other Business

  • Peter: Any thoughts on Developer Day or the hackathon from those that attended?
    • Ben: Developer Day went really well. Hackathon would have been better with a bit more structure to it with having working groups set up in advance.
    • Peter: I agree and we thought we had this set up in advance, but we didn’t execute well on it at the event.
    • Cory: I thought hackathon was pretty cool and maybe we should have devoted my time to it. Once we started people wanted to go off and join the social event.
    • Ken: The time pressure issue could be fixed, maybe post-LUG event with RSVP and topics set up in advance would be better.

Next meeting will be on 2017-06-29 at 11:00am Pacific