You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-issues@hadoop.apache.org by "Hudson (JIRA)" <ji...@apache.org> on 2018/04/24 20:59:06 UTC
[jira] [Commented] (HDFS-12864) Ozone: Turn off full synced write
in RocksDB MetadataStore
[ https://issues.apache.org/jira/browse/HDFS-12864?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16451124#comment-16451124 ]
Hudson commented on HDFS-12864:
-------------------------------
SUCCESS: Integrated in Jenkins build Hadoop-trunk-Commit #14057 (See [https://builds.apache.org/job/Hadoop-trunk-Commit/14057/])
HDFS-12864. Ozone: Turn off full synced write in RocksDB MetadataStore. (aengineer: rev ffccb190074782ef5d2b896063639c4de62aca84)
* (edit) hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/utils/RocksDBStore.java
> Ozone: Turn off full synced write in RocksDB MetadataStore
> ----------------------------------------------------------
>
> Key: HDFS-12864
> URL: https://issues.apache.org/jira/browse/HDFS-12864
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ozone
> Reporter: Elek, Marton
> Assignee: Elek, Marton
> Priority: Major
> Fix For: HDFS-7240
>
> Attachments: HDFS-12864-HDFS-7240.001.patch
>
>
> Based on the suggestion from [~xyao] I created the patch to turn off fully synced write in the rocksdb.
> With this patch we measured about 10x performance gain (200 nodes cluster + 30 corona instance).
> 1. We don't need fully synced rocksdb writing, as HA will cover the few cases when a node would be lost
> 2. Long term (IMHO) we need to add a possibility to configure any settings in rocksdb, so it could be turned off, but the setSync(false) is a reasonable default.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org