You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Yun Tang (JIRA)" <ji...@apache.org> on 2018/12/13 03:39:00 UTC

[jira] [Commented] (FLINK-11138) RocksDB's wal seems to be useless in rocksdb-state-backend that reduce the disk's overhead

    [ https://issues.apache.org/jira/browse/FLINK-11138?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16719713#comment-16719713 ] 

Yun Tang commented on FLINK-11138:
----------------------------------

Yes, you are right. But we already disable the WAL of rocksDB. You could view configuration  [here|https://github.com/apache/flink/blob/967b31b333e6f4b014ea3041f420bfaff2484618/flink-state-backends/flink-statebackend-rocksdb/src/main/java/org/apache/flink/contrib/streaming/state/RocksDBKeyedStateBackend.java#L327].

> RocksDB's wal seems to be useless in rocksdb-state-backend that reduce the disk's overhead
> ------------------------------------------------------------------------------------------
>
>                 Key: FLINK-11138
>                 URL: https://issues.apache.org/jira/browse/FLINK-11138
>             Project: Flink
>          Issue Type: Improvement
>            Reporter: cailiuyang
>            Priority: Major
>
> The wal of rocksdb is used to recovery after crash, but after flink-job crash, it always recovery from last checkpoint, so it seems wal is useless.
> Am i right?



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)