feat(server): flush slots traverse better yield #4821
Conversation
Signed-off-by: adi_holden <[email protected]>
src/server/db_slice.cc
Outdated
| int64_t exec_time_ms = (after - start) / 1000000; | ||
| if (exec_time_ms > 10) { |
There was a problem hiding this comment.
10ms is a lot, imho. why not reduce to 1?
There was a problem hiding this comment.
ok we can do 1ms
There was a problem hiding this comment.
@romange do we have anything cheaper than GetCurrentTimeNanos(), because we need ms precision
There was a problem hiding this comment.
i think GetCurrentTimeNanos() is ok here, given that we do not run the function frequently.
src/server/db_slice.cc
Outdated
| if (i % 100 == 0) { | ||
| int64_t after = absl::GetCurrentTimeNanos(); | ||
| int64_t exec_time_ms = (after - start) / 1000000; | ||
| if (exec_time_ms > 10) { |
There was a problem hiding this comment.
Since we use micro in all of our statistics maybe worth keeping that unit ?
src/server/db_slice.cc
Outdated
| auto iterate_bucket = [&](PrimeTable::bucket_iterator it) { | ||
| it.AdvanceIfNotOccupied(); | ||
| while (!it.is_done()) { | ||
| del_entry_cb(it); |
There was a problem hiding this comment.
del_entry_cb is redundant now. why not move its implementation here?
There was a problem hiding this comment.
you are right, this was just easy for me to test this with moving back and forward with TraverseBuckets/Traverse to see the impact of this change
src/server/db_slice.cc
Outdated
| @@ -787,13 +787,16 @@ void DbSlice::Del(Context cntx, Iterator it) { | |||
| } | |||
|
|
|||
| void DbSlice::FlushSlotsFb(const cluster::SlotSet& slot_ids) { | |||
| VLOG(3) << "Start FlushSlotsFb"; | |||
There was a problem hiding this comment.
I suggest making it 2 or even 1, because it's not often operation
Signed-off-by: adi_holden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
Signed-off-by: adi_holden <[email protected]>
No description provided.