[LU-3978] Typos in NRS ORR/TRR policy completely invalidate scheduling behaviour Created: 20/Sep/13 Updated: 30/Oct/13 Resolved: 23/Sep/13 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | None |
| Fix Version/s: | Lustre 2.5.0, Lustre 2.4.2 |
| Type: | Bug | Priority: | Major |
| Reporter: | Nikitas Angelinas | Assignee: | Emoly Liu |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | patch | ||
| Severity: | 3 |
| Rank (Obsolete): | 10613 |
| Description |
|
Some CRR-N policy fields are used by the ORR/TRR policies in lustre/ptlrpc/nrs_orr.c; this is the result of a repeated typo, but completely impairs the ability of ORR/TRR to perform any sensible scheduling. Sadly, this would mean that any performance tests that may have been carried out using these policies, since approximately the feature freeze date for 2.4, would have to be repeated if one is interested in obtaining useful benchmarking results. Sorry, a lower case 'c' can look a lot like a lower case 'o' in an editor window, when using these small font sizes, especially after a few hours of staring at the monitor. |
| Comments |
| Comment by Nikitas Angelinas [ 20/Sep/13 ] |
|
patch for master is at http://review.whamcloud.com/#/c/7708 |
| Comment by Peter Jones [ 20/Sep/13 ] |
|
Nikitas I conçur and am glad that you have now caught this Emoly Could you please take care of this one? Thanks Peter |
| Comment by Andreas Dilger [ 20/Sep/13 ] |
|
I'm actually glad to see something like this, because the ORR test results were completely underwhelming. I'd be happy to see new performance benchmarks, particularly with ZFS, to see if this actually improves IO performance. |
| Comment by Jodi Levi (Inactive) [ 23/Sep/13 ] |
|
Patch landed to Master so closing ticket. |