LWG Minutes 2018-03-08

Attendance
Cray: Cory Spitz, Ben Evans

ORNL: James Simmons

HPE: Olaf Weber, 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


 * Oleg just did another batch of landings today that cleared off a number of the remaining tickets that were still marked in scope.
 * We have discovered a new issue due to an earlier landing.
 * Andreas: We recently landed a patch to make the default stripe size 4MB which shows a reasonable performance improvement.  It had some issues with testing assuming that the stripe size was 1MB, however, now we see that when this is combined combined with DoM it presents an issue where users could fill up their MDT unintentionally.  We need to fix the 1MB DoM / 4MB interaction, which is a significant change that needs to be well thought out.  To resolve the issue, we think we should revert the change to make 4MB the default until we can come up with a proper solution to handle the DoM interaction better.
 * Does anyone have any feedback with testing or otherwise?
 * James: I will be testing changelogs very soon, in the next couple of days.  I will be setting up cray system to test as well.
 * Cory: We are currently doing some small scale testing.  The good news is that we haven’t identified any serious issues.  We hope to get scale testing going in the next few days.
 * Cory: LU-5216.  We would like to get it fixed for 2.11 if possible.
 * Peter: John was looking into this and thought it would need some work to get done.
 * Ben: LU-5216 is a bug we are seeing in the field.  Getting it landed and then pulled back if it is only due to coding standards is a little frustrating if it can’t be fixed for this release.
 * Peter: We agree that we need to fix it, but it really doesn’t have to land for this particular release and we should look at other paths forward.  If you have a fix that works for your customers, nothing would prevent from continuing to use it until a more appropriate fix can be landed for all to use.
 * Cory: That makes sense and we will go back and work on it for 2.12 and we will consult with John.
 * Abe: We may be able to test a bit next week.

Upstream Lustre Client

James


 * Thank you to Andreas for reviewing patches that touch very old areas of code and few people have expertise in those areas.
 * mds-survey hangs on the upstream client with the latest master (sanity test 225).
 * Andreas: This is only meant to be run on the server and likely would be the cause of issues.

lustre.org

Ken (via email)


 * We are in the process of another round of security updates. The hope is to have those completed this weekend.
 * Research into the possibility of using Stack Overflow or a Stack Exchange community site for Lustre Q/A has been promising. Initial indications are that this is workable and that generated content there should be compatible with our licensing guidelines. We'll be following up further with Stack Exchange and investigating how to potentially integrate this with our existing web resources.

Other Business


 * Developer Day
 * Peter: I am hoping to have a straw man out for developer day in the coming weeks.
 * Peter: James, is an hour enough for you to go over upstream work?
 * James: Yes, that should be fine.  I would like to cover udev and a few other things.
 * James: Does anyone know if Neil Brown is coming?
 * Peter: I am not aware of his status, but it could be useful to have him present at such an event.

Next meeting will be on 2018-03-22 at 11:00am Pacific