Add debugging log for index cp. #592
Conversation
|
Codecov ReportAttention: Patch coverage is
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master #592 +/- ##
===========================================
+ Coverage 56.51% 66.62% +10.10%
===========================================
Files 108 109 +1
Lines 10300 10774 +474
Branches 1402 1473 +71
===========================================
+ Hits 5821 7178 +1357
+ Misses 3894 2889 -1005
- Partials 585 707 +122 ☔ View full report in Codecov by Sentry. |
xiaoxichen
force-pushed
the
debug_log
branch
from
c1bf95d to
262f94e
Compare
| @@ -89,7 +89,7 @@ void IndexService::start() { | |||
| } | |||
| // Force taking cp after recovery done. This makes sure that the index table is in consistent state and dirty buffer | |||
| // after recovery can be added to dirty list for flushing in the new cp | |||
| hs()->cp_mgr().trigger_cp_flush(true /* force */); | |||
| hs()->cp_mgr().trigger_cp_flush(true /* force */).wait(); | |||
There was a problem hiding this comment.
trigger_cp_flush(true) waits until cp flush is done. Did you test it without .wait()
There was a problem hiding this comment.
Right now all the tests are done without the wait()... I dont know if we do need a wait() but it seems make things easier...
There was a problem hiding this comment.
wait() will guarantee the cp superblk is written to disk, which further means the last_cp_id is persisted. if this does not do any harm to performance, I think we can wait here, or even all the trigger_cp_flush
Signed-off-by: Xiaoxi Chen <[email protected]>
xiaoxichen
force-pushed
the
debug_log
branch
from
262f94e to
ac79b6c
Compare
|
@xiaoxichen do you have a plan to merge this pr or close it? |
No description provided.