hangc0276 opened a new pull request, #3646:
URL: https://github.com/apache/bookkeeper/pull/3646

   ### Motivation
   When there are a lot of ledgers to be deleted, the EntyLocationIndex in 
RocksDB will delete a large number of entries at one time. 
   ```
   Nov 11, 2022 @ 05:49:39.336     05:49:39.336 [db-storage-cleanup-10-1] INFO  
org.apache.bookkeeper.bookie.storage.ldb.EntryLocationIndex - Deleted indexes 
for 0 entries from 16074 ledgers in 625.242 seconds
   Nov 11, 2022 @ 10:18:08.281     10:18:08.281 [db-storage-cleanup-10-1] INFO  
org.apache.bookkeeper.bookie.storage.ldb.EntryLocationIndex - Deleted indexes 
for 0 entries from 13376 ledgers in 248.66 seconds
   Nov 11, 2022 @ 14:27:06.949     14:27:06.949 [db-storage-cleanup-10-1] INFO  
org.apache.bookkeeper.bookie.storage.ldb.EntryLocationIndex - Deleted indexes 
for 0 entries from 11798 ledgers in 265.118 seconds
   Nov 11, 2022 @ 19:45:40.593     19:45:40.593 [db-storage-cleanup-10-1] INFO  
org.apache.bookkeeper.bookie.storage.ldb.EntryLocationIndex - Deleted indexes 
for 0 entries from 15546 ledgers in 1133.236 seconds
   ```
   The RocksDB location index deletes operation will cause RocksDB busy, and 
impact the read latency to a few seconds.
   
   Currently, we delete RocksDB entries in batch, and the batch size is 
hard-coded. We can not change it.
   
   ### Changes
   Make the RocksDB delete entries in batch configurable.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to