You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "刘方奇 (Jira)" <ji...@apache.org> on 2021/12/04 01:00:34 UTC

[jira] [Created] (FLINK-25163) Add more options for rocksdb state backend to make configuration more flexible

刘方奇 created FLINK-25163:
---------------------------

             Summary: Add more options for rocksdb state backend to make configuration more flexible
                 Key: FLINK-25163
                 URL: https://issues.apache.org/jira/browse/FLINK-25163
             Project: Flink
          Issue Type: Improvement
          Components: Runtime / State Backends
            Reporter: 刘方奇


Now flink has less options than the configurations what Rocksdb can set. We can see many function in the org.rocksdb.DBOptions that can influence its behavior(e.g rocksdb background threads).

It make us do less when we want to do some thing to tuning Rocksdb. In my opinion, there are at least there options:
 # maxBackgroundFlushes, it can define the background flush threads. default 1.
 # maxBackgroundCompactions, it can define the background compaction threads. default 1.
 # maxBackgroundJobs, it can define the background threads. default 2.

setIncreaseParallelism (the most we can do for the rocksdb backend background threads) seems like can do little. It can't change the flush threads. I think it's necessary to make rocksdb configuration flexible.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)