LWG Minutes 2019-03-21
Cray: Cory Spitz
ORNL: Dustin Leverman, James Simmons
HPE: Olaf Weber, Christopher Voltz
IU: Ken Rawlings
Sandia: Ruth Klundt
Kmesh: Michael Nishimoto
TACC: Ari Martinez
LLNL: Olaf Faaland
Whamcloud: Joe Gmitter, Peter Jones, Andreas Dilger, James Nunez
New Actions Captured:
Existing Open Actions:
Actions Recently Closed:
2.13.0 Release Update
- We have been continuing with landings. We are getting close to the feature cutoff. Are there any updates to the likelihood of features making it in time?
- Joe: PCC is mainly already in flight on the topic branch. We need to do some trimming of what is going to be targeted to land and will have those discussions ahead of the next call. UDSP is a bit more risk as some of the same resources are working on sysfs, however, the team still feels confident in getting all patches into Gerrit on schedule.
- Cory: At the moment, the end of March is not looking too good. We have reasoned the SEL ticket to Vitaly. He is reviewing the past work and has some changes he has to push ahead of the cutoff.
- James: I have a netlink base API patch and not getting inspections, will try to reach out to get more interest in looking at it.
Upstream Lustre Client
- Things are going well and James has been reviewing Neil's patches.
- Neil is looking at the openSFS tree as well.
- We will be proposing to do a move out of staging, but have been distracted with other issues on master at the moment.
Ken (via email)
- The PHP upgrade happened last week as scheduled and we have been testing. Please let us know if anyone sees any issues.
Next up is the HTTPS issue discussed last time and we are also working on the support matrix discussed last call.
- LNet/Multi-Rail testing discussion proposed by Cory
- Peter. We have static analysis that runs on every patch submitted and are growing the LUTF to be able to run via AT for more complete LNet testing.
- Cory: We are going to be putting together a system to do some testing of MR configuration. We have tried using the LUTF, but it doesn't seem quite there yet.
- Andreas: We have taken on the testing voluntarily, so it wouldn’t be terrible if someone else contributed to that effort. Andreas suggested that if Cray has an AutoTest equivalent, that a bot could run on a patch and comments to Gerrit for additional testing feedback.
- Peter: You could also react to landings, test, and then comment back.
- Cory: Maybe it is better for us to test master-next instead of master to catch things earlier.
- Peter: Ideally yes, however, master-next is only around for a short period of time so that needs to be considered.
Next meeting will be on 2019-04-04 at 11:00am Pacific