[
https://issues.apache.org/jira/browse/FLINK-9070?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16479058#comment-16479058
]
ASF GitHub Bot commented on FLINK-9070:
---------------------------------------
Github user sihuazhou commented on the issue:
https://github.com/apache/flink/pull/5979
@StefanRRichter , the reason I prefer this approach is that:
- From the comment in RocksDB's source we can find that deleteRange()
should be used for deleting big range, what if the entries num of the map is
not that big.
- From the comments we can also find that deleteRange() would hurt the read
performance, so we should consider to set ReadOptions::ignore_range_deletions =
true to avoid the negative effect by deleteRange(), but if we use it for
MapState.clear(), it seems that we can't set
ReadOptions::ignore_range_deletions = true.
And current approach should not bring any downside, what do you think?
> Improve performance of RocksDBMapState.clear()
> ----------------------------------------------
>
> Key: FLINK-9070
> URL: https://issues.apache.org/jira/browse/FLINK-9070
> Project: Flink
> Issue Type: Improvement
> Components: State Backends, Checkpointing
> Affects Versions: 1.6.0
> Reporter: Truong Duc Kien
> Assignee: Sihua Zhou
> Priority: Major
> Fix For: 1.6.0
>
>
> Currently, RocksDBMapState.clear() is implemented by iterating over all the
> keys and drop them one by one. This iteration can be quite slow with:
> * Large maps
> * High-churn maps with a lot of tombstones
> There are a few methods to speed-up deletion for a range of keys, each with
> their own caveats:
> * DeleteRange: still experimental, likely buggy
> * DeleteFilesInRange + CompactRange: only good for large ranges
>
> Flink can also keep a list of inserted keys in-memory, then directly delete
> them without having to iterate over the Rocksdb database again.
>
> Reference:
> * [RocksDB article about range
> deletion|https://github.com/facebook/rocksdb/wiki/Delete-A-Range-Of-Keys]
> * [Bug in DeleteRange|https://pingcap.com/blog/2017-09-08-rocksdbbug]
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
