You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Aljoscha Krettek (JIRA)" <ji...@apache.org> on 2016/04/20 16:17:25 UTC
[jira] [Resolved] (FLINK-3718) Add Option For Completely Async
Backup in RocksDB State Backend
[ https://issues.apache.org/jira/browse/FLINK-3718?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Aljoscha Krettek resolved FLINK-3718.
-------------------------------------
Resolution: Fixed
Fix Version/s: 1.1.0
Implemented in https://github.com/apache/flink/commit/c33325e3ffb9acecfc9c562b21a3ac703bae983d
> Add Option For Completely Async Backup in RocksDB State Backend
> ---------------------------------------------------------------
>
> Key: FLINK-3718
> URL: https://issues.apache.org/jira/browse/FLINK-3718
> Project: Flink
> Issue Type: Bug
> Components: Streaming
> Reporter: Aljoscha Krettek
> Assignee: Aljoscha Krettek
> Fix For: 1.1.0
>
>
> Right now, the snapshotting for RocksDB has a synchronous part where a backup of the RocksDB database is drawn and an asynchronous part where this backup is written to HDFS.
> We should add an option that uses the snapshot feature of RocksDB to get an iterator over all keys at a set point in time. The iterator can be used to store everything to HDFS. Normal operation can continue while we store the keys. This makes the snapshot completely asynchronous.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)