Omap small bugs #6046
Omap small bugs #6046
Conversation
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
When set omapheader/omapvals, it alreade set FLAG_OMAP. So for remove don't set again. Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
The code already check object whether exist. So it don't need call touch. Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
…deep-scrub. Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
For be_deep_scrub, we already got the hash info attr. So if don't found in cache, directly decode rather than call stat & getattr. Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
Now erasure pool don't support omap. So it dont need set those. Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
… omap. Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
|
@athanatos . Have you time to review those? thanks |
| @@ -409,7 +409,8 @@ enum scrub_error_type PGBackend::be_compare_scrub_objects( | |||
| if (auth.size != candidate.size) { | |||
| if (error != CLEAN) | |||
| errorstream << ", "; | |||
| error = SHALLOW_ERROR; | |||
| if (error != DEEP_ERROR) | |||
| error = SHALLOW_ERROR; | |||
There was a problem hiding this comment.
There is a comment above that says that the shallow error here takes precedence because this can be seen by both kinds of scrubs. This is a bit odd, but I'd say remove this entire commit unless you are seeing a particular problem.
There was a problem hiding this comment.
Hmm... making shallow take precedence doesn't make sense to me. Anybody remember why it says that? @athanatos ?
There was a problem hiding this comment.
@liewegas In addition to your change at line 412 we should not set read_error in be_scan_list() on stat() errors and set the stat size to something invalid instead. Also, remove the comments at line 374 and at line 407.
This way a regular scrub will never get a DEEP_ERROR from this function, but a deep-scrub could possibly see a SHALLOW_ERROR if only stat() fails. This keeps the error counts from getting messed up.
ghost
added
|
@tchaikov . Have you time to review this? Thanks! |
|
This needs rebase. And a run through the qa suite. (Ping us in #sepia if you need help using teuthology-openstack!) |
|
I've got a rebase and includes addition of stat_error as pull request #6669 |
|
reviewed. |
|
needs rebase? |
|
nm, see @dzafman rebase |
No description provided.