CDWG Minutes 2013-01-30
- Nic Henke (Cray)
- Andreas Dilger (Intel)
- Peter Jones (Intel)
- Bryon Neitzel (Intel)
- Jodi Levi (Intel)
- James Simmons (ORNL)
- Chris Morrone (LLNL)
Chris suggested providing more granularity for distributions (instead of "RHEL 6.X", do "RHEL 6.1", 6.2, 6.3, etc.). Might be nice to know, but would make a very crowded set of radio boxes. Peter will consider it, but can stick to current question style if it would make the survey too cluttered.
No other comments.
Lustre 2.4 feature freeze update
Peter reports that the following features are likely to land in time for 2.4:
- NRS (Network Request Scheduler)
- 4MB I/O support
These features are close, but there landing is not yet definite (and will require at least a few more days):
- HSM - full HSM will not be in 2.4, but 2.4 client support should land
- DNE phase 1
- Tool for migration from 1.8
Nick requested updates on the CDWG list and/or other major list Lustre mailing list.
ORNL pre-2.4 test shot on Jaguar
James discussed the results of the recent tests of the code from master/pre-2.4 on Jaguar.
It went "more smoothly than we've experienced in the past"
Only one job didn't complete, and probably legitimately ran out of disk space
Noted that Cray 1.8 clients did not work with the new servers
Tests included, ior, mdtest, s3d
James asked if LU-1812 (Fedora 18 support) could still land after the "feature freeze" date.
Peter said that LU-1812 is considered a "minor enhancement", and since it is well on its way to completion and has little to no impact on the rest of the code, it is not constrained by the "feature freeze" date.
Andreas asked about the two competing FC18/SP2 patche series. James reported that they are working together and have consolidated most of their work. Some patches for ldiskfs not yet merged, but in the works.
James asked about patchless server for 2.4. We are close to patchless, but it looks like 2.4 will not quite eliminate the patches. Therefore the FC18/SP2 patches should be included.
Andreas mentioned that Xyratex submitted mdraid patches to add to the kernel patch series. Those have a good chance to land for 2.4. When we go patchless, they'll need to move somewhere else. Would be nice if Xyratex can get those patches landed upstream.