[
https://issues.apache.org/jira/browse/HBASE-5954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14222311#comment-14222311
]
Sean Busbey commented on HBASE-5954:
------------------------------------
Why support the mix though?
The last time we checked the relative cost of hflush vs hsync over in Accumulo
it was a severe penalty. Once we're in a world of prevalent SSDs and hsync is
cheaper, why wouldn't we do it always, atleast for a given cluster?
I'm just wondering how many different practical use cases we gain by allowing
the flexibility of per-request no-wal vs flush vs sync compared to a
cluster-wide setting of flush vs sync and then per-request no-wal or whichever
is configured.
Since this would presumably be a configurable property on the wal provider, we
could still make meta always hsync.
> Allow proper fsync support for HBase
> ------------------------------------
>
> Key: HBASE-5954
> URL: https://issues.apache.org/jira/browse/HBASE-5954
> Project: HBase
> Issue Type: Improvement
> Components: HFile, wal
> Reporter: Lars Hofhansl
> Assignee: Lars Hofhansl
> Priority: Critical
> Fix For: 2.0.0
>
> Attachments: 5954-WIP-trunk.txt, 5954-trunk-hdfs-trunk-v2.txt,
> 5954-trunk-hdfs-trunk-v3.txt, 5954-trunk-hdfs-trunk-v4.txt,
> 5954-trunk-hdfs-trunk-v5.txt, 5954-trunk-hdfs-trunk-v6.txt,
> 5954-trunk-hdfs-trunk.txt, hbase-hdfs-744.txt
>
>
> At least get recommendation into 0.96 doc and some numbers running w/ this
> hdfs feature enabled.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
