osd/OSDMap: improve the performance of pg_to_acting_osds #12190
Conversation
|
The case is: when pg temp is not empty, acting will be calculated without ups and up_primary. But ups and up_primary are always calculated, and affect the performance at this situation. |
|
I find one thing very weird. I created this pr yesterday morning, and found one testcase failed when ran jekins: So I had to close this pr, and investigated what was wrong, but nothing found. When I reopened this pr without any changes, jekins passed.... Anyone could give me any idea? |
liewegas
changed the title
|
@liupan1111 there are still some issues with the unit tests; it sounds like it wasn't caused by this change. |
| _raw_to_up_osds(*pool, raw, &_up, &_up_primary); | ||
| _apply_primary_affinity(pps, *pool, &_up, &_up_primary); | ||
| if (_acting.empty()) { | ||
| _acting = _up; |
There was a problem hiding this comment.
and could you fix the Signed-off-by line in your commit message? or just squash these two commits into a single one?
| _acting = _up; | ||
| if (_acting_primary == -1) { | ||
| _acting_primary = _up_primary; | ||
| } | ||
| } | ||
| } | ||
| if (up) |
There was a problem hiding this comment.
you could mvoe the if (up) and if (up_prmiary) lines into the block?
|
@liewegas , modified as your request. |
| _raw_to_up_osds(*pool, raw, &_up, &_up_primary); | ||
| _apply_primary_affinity(pps, *pool, &_up, &_up_primary); | ||
| if (_acting.empty()) { | ||
| _acting = _up; |
There was a problem hiding this comment.
@liupan1111 could you fix the indent?
|
@tchaikov done, thanks! |
|
@liupan1111 could you fix the signed-off-by line in 993d37d, or better off squashing these two commit into a single one? |
|
@tchaikov what is wrong with signed-off-by? |
- Signed-off-by: Pan Liu pan.liu@istuary.com
+ Signed-off-by: Pan Liu <pan.liu@istuary.com> |
Signed-off-by: Pan Liu <pan.liu@istuary.com>
|
@tchaikov modified. |
…up vector" is not requested.
Signed-off-by: Pan Liu pan.liu@istuary.com