LWG Minutes 2019-06-27: Difference between revisions

From OpenSFS Wiki
Jump to navigation Jump to search
(Created page with "== Attendance == CEA: Jacques-Charles Lafoucriere <br /> Cray: Cory Spitz <br /> HPE: Olaf Weber, Christopher Voltz, Jeff Garlough <br /> IU: Ken Rawlings <br /> ORNL: Dus...")
 
No edit summary
Line 1: Line 1:
== Attendance ==
== Attendance ==
CEAJacques-Charles Lafoucriere <br />
ORNLDustin Leverman <br />
Cray: Cory Spitz <br />
Cray: Cory Spitz <br />
HPE:  Olaf Weber, Christopher Voltz, Jeff Garlough <br />
HPE:  Olaf Weber, Christopher Voltz, Jeff Garlough <br />
IU:  Ken Rawlings <br />
IU:  Ken Rawlings <br />
ORNLDustin Leverman, James Simmons <br />
KmeshMichael Nishimoto <br />
TACC:  Ari Martinez <br />
TACC:  Ari Martinez <br />
Whamcloud:  Peter Jones, James Nunez, Andreas Dilger, Patrick Farrell <br />
Whamcloud:  Peter Jones, Joe Gmitter, Andreas Dilger <br />


== Actions ==
== Actions ==
Line 20: Line 20:
'''2.13.0 Release Update''' <br />
'''2.13.0 Release Update''' <br />
Peter <br />
Peter <br />
*The Self-Extending Layout still not yet landed but is looking close. Alexey Lyashkov (Cray) will review the latest versions when he is back from vacation on Monday. Further testing occurring to reassure that higher than average observation of test failures is not triggered by these patches
*There has been a lot of activity around self-extending layouts in master-next and things are looking good so far. It should land today and this will mark us being feature complete for this release.
*Cray will handle testing of SEL on master prior to 2.13 GA
*James mentioned Arm server support. ORNL management has asked for him to be responsible for ongoing vetting of testing results so that we can introduce this
<br />
<br />


'''Upstream Lustre Client''' <br />
James <br />
*Working on 2.11 patches
*Neil on vacation for next two weeks
<br />


'''lustre.org''' <br />
'''lustre.org''' <br />
Ken <br />
Ken <br />
*Upgrading to Media wiki 1.3.0
*We are continuing to upgrade quite a bit on the lustre.org main site.  We probably have a few more days of updates to be 100% complete.
*If anyone runs into any subtle issues with the site, please let Ken know.
*We are looking to do the mediawiki update in the next few weeks as well.
<br />
<br />


'''Other Business''' <br />
'''Other Business''' <br />
*James is concerned that there are still issues with changelogs
*Dustin:  We recently upgraded to 2.12.2 and, as part of that, we upgraded to zfs 0.8.1.  We ran into a dnode cache bug and it looks like other sites are also reporting it.  We are considering reverting to 2.11 and zfs 0.7.13.
*Peter suggested thinking ahead for 2.15 features for the roadmap to discuss next call
**Peter:  Have you tried 2.12.2 with zfs 0.7.13?  You don't need to revert the Lustre version to roll back the zfs version.
**Dustin:  Agreed that it is not a lustre issue, but since the customer is upset we are just going back to old image.
**Andreas:  We didn’t do much with 0.8.X and 2.12 since it was not released yet at the time of our testing.
**The issue is captured in https://jira.whamcloud.com/browse/LU-12510.
*Roadmap Discussion
**Andreas:  We may be able to put on immediate write resync, but that’s post 2.15.
**Cory:  In the 2.15 timeframe for Cray we are thinking of working on the hsm coordinator infrastructure, update to file copy tools to get it working between FLR and HSM.
**Cory:  We should probably update project page as part of the roadmap update, it is quite stale.
**Cory:  We would like to collaborate around metadata write back cache if possible, there would be value there on our side as well.
**Olaf:  James is concerned that there is still issue with changelongs and he is not the only one (LU-11426, LU-11205, LU-11581).
 
<br />
<br />


'''Next meeting will be on 2019-07-11 at 11:00am Pacific'''
'''Next meeting will be on 2019-07-25 at 11:00am Pacific'''

Revision as of 07:46, 17 July 2019

Attendance

ORNL: Dustin Leverman
Cray: Cory Spitz
HPE: Olaf Weber, Christopher Voltz, Jeff Garlough
IU: Ken Rawlings
Kmesh: Michael Nishimoto
TACC: Ari Martinez
Whamcloud: Peter Jones, Joe Gmitter, Andreas Dilger

Actions

New Actions Captured:

  • None

Existing Open Actions:

  • None

Actions Recently Closed:

  • Peter sent out roadmap slide ahead of ISC

Minutes

2.13.0 Release Update
Peter

  • There has been a lot of activity around self-extending layouts in master-next and things are looking good so far. It should land today and this will mark us being feature complete for this release.



lustre.org
Ken

  • We are continuing to upgrade quite a bit on the lustre.org main site. We probably have a few more days of updates to be 100% complete.
  • If anyone runs into any subtle issues with the site, please let Ken know.
  • We are looking to do the mediawiki update in the next few weeks as well.


Other Business

  • Dustin: We recently upgraded to 2.12.2 and, as part of that, we upgraded to zfs 0.8.1. We ran into a dnode cache bug and it looks like other sites are also reporting it. We are considering reverting to 2.11 and zfs 0.7.13.
    • Peter: Have you tried 2.12.2 with zfs 0.7.13? You don't need to revert the Lustre version to roll back the zfs version.
    • Dustin: Agreed that it is not a lustre issue, but since the customer is upset we are just going back to old image.
    • Andreas: We didn’t do much with 0.8.X and 2.12 since it was not released yet at the time of our testing.
    • The issue is captured in https://jira.whamcloud.com/browse/LU-12510.
  • Roadmap Discussion
    • Andreas: We may be able to put on immediate write resync, but that’s post 2.15.
    • Cory: In the 2.15 timeframe for Cray we are thinking of working on the hsm coordinator infrastructure, update to file copy tools to get it working between FLR and HSM.
    • Cory: We should probably update project page as part of the roadmap update, it is quite stale.
    • Cory: We would like to collaborate around metadata write back cache if possible, there would be value there on our side as well.
    • Olaf: James is concerned that there is still issue with changelongs and he is not the only one (LU-11426, LU-11205, LU-11581).


Next meeting will be on 2019-07-25 at 11:00am Pacific