fix log flush bug #460
fix log flush bug #460
Conversation
| @@ -422,7 +420,7 @@ void HomeLogStore::flush_sync(logstore_seq_num_t upto_seq_num) { | |||
| m_logdev->flush_if_needed(1); | |||
|
|
|||
| // Step 4: Wait for completion | |||
| m_sync_flush_cv.wait(lk, [this, upto_seq_num] { return !m_records.status(upto_seq_num).is_active; }); | |||
There was a problem hiding this comment.
I am not sure if we will have multiple concurrent flush_sync , if we have, this implementation is not right. Although the previous one is also not right:)
There was a problem hiding this comment.
Not necessary concurrent...
First call with flush_sync (100), it execute to L416. At the moment, the on_write_completion run to finish and set the m_sync_flush_promise. The flush_sync (100) fulfilled in L417 and not consume the future.
Next time , a flush_sync (1078576) will return immediately by consuming the previous fulfilled future.
There was a problem hiding this comment.
do you mean there is a potential bug？ this part is protected by a lock, and everytime we enter this block, m_sync_flush_promise will be updated to a new one.
There was a problem hiding this comment.
you can try it locally in your env. it runs well for hours in mine
|
will use sync flush , so this pr is unnecessary again |
No description provided.