LWG Minutes 2018-03-22: Difference between revisions

From OpenSFS Wiki
Jump to navigation Jump to search
(Created page with "== Attendance == Cray: Ben Evans <br /> ORNL: James Simmons <br /> HPE: Christopher Voltz, Robert Clark, Stephen Champion <br /> Sandia: Ruth Klundt <br /> SuperMicro: Ab...")
 
mNo edit summary
 
Line 66: Line 66:


'''Next meeting will be on 2018-04-05 at 11:00am Pacific'''
'''Next meeting will be on 2018-04-05 at 11:00am Pacific'''
[[Category:LWG]]

Latest revision as of 05:40, 8 April 2019

Attendance

Cray: Ben Evans
ORNL: James Simmons
HPE: Christopher Voltz, Robert Clark, Stephen Champion
Sandia: Ruth Klundt
SuperMicro: Abe Asraoui
Intel: Joe Gmitter, Peter Jones, Andreas Dilger

Actions

New Actions Captured:

  • None

Existing Open Actions:

  • None

Actions Recently Closed:

  • None

Minutes

2.11.0 Release Update
Peter

  • RC1 is in testing and looking pretty good. We expect to complete the first long pass through RC1 next week.
  • We have found one issue that affects Ubuntu clients, where a tweak is needed on the build side. We will be putting up an RC2 with those build changes.
  • Ben, can you talk to any Cray testing for 2.11.0?
    • Ben:We ran into LU-9699, but not sure what else has been found.
    • Peter: Have you hit this in scale testing or is it orthogonal to that?
      • Ben: It is something that we hit ever time a reconfig happens with writeconf.
      • Peter: So this is not something specific to the scale testing, but rather something that has been existing for some time. This seems like something that we would not hold the release for at this point since all releases with DNE would have this issue present. We can fix it in an LTS release when ready and the fix would then be available for anyone hitting it on master as well.
  • James, do you have any testing feedback?
    • James: I got the ARM system and Power8 functional. Things are looking pretty good. Still seeing some funky behavior with changelogs, but I haven’t been able to look into it further.
    • James: I think I only see it in a DNE setup, where of the MDTs never recovers. I will file a ticket with details after collecting some logs.
  • Anyone else doing any testing?
    • Robert: We are getting ready to do some 2.11 stuff m, but nothing active at the moment.
    • Abe: Nothing yet, hopefully by the end of the month we will be able to start some 2.11 testing.


Upstream Lustre Client
James

  • Doug Oucharek has been doing some infiniband patches.
  • Completed a number of SMP changes and should be fully completed soon.
  • Have also been testing sysfs patches and working with John Hammond on an xattr bug.
  • There is lots on the boiler plate to push upstream soon.
  • Jinshan has been contact James about the loopback driver. James wants to talk with him and Neil further about this to avoid rewriting our own again.
  • Dmitry has been back porting some fixes back to master as well.


lustre.org
Ken (via email)

  • Shawn Hall and I are working through procedural and technical aspects on the Q/A site question and the potential integration with Stack Overflow and will have an update on the next LWG call.
  • The original plan for migrating the remaining old wiki pages proved unworkable. However, MediaWiki's collapsible pages functionality is looking promising. We'll be testing that for the remaining pages on a staging server and if successful that will allow us to finish the migration and move on to planning & scheduling the old wiki retirement.


Other Business

  • LUG
    • The agenda for the main conference was published yesterday.
    • Developer Day
      • Peter: We have made some further progress. So far I know of 4 engineers each from Cray, DDN, and Intel coming, as well as one each from LLNL, ORNL, and Sandia. We are waiting to hear back on a few more.
      • We hope to have an agenda by next week.
  • Roadmap
    • We are trying to get the Lustre roadmap refreshed ahead of LUG.
    • We are refreshing a few things on the Intel side. There are a few more LTS releases to note, we need to plan out FLR across releases, add the 2.11 date, and extend the time window by 6 months or so.
    • We are thinking of showing the releases as a 7 month cycle to be more accurate of what we see in actuality.
      • James: It sounds like we will just push more content in and be in the same situation.
      • Peter: It seems that features tend not to be the issues that causes problems in stabilization. It has been bug fixes or fixes that are refactoring code to match upstream. Those types of fixes touch areas of the code base that is sensitive and end up being found during stabilization.


Next meeting will be on 2018-04-05 at 11:00am Pacific