[jira] [Commented] (HBASE-19024) Configurable default durability for synchronous WAL

["Yu Li (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-19024?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16445592#comment-16445592 ] 

Yu Li commented on HBASE-19024:
-------------------------------

Sorry to be late for the party, just noticed this one.

Skimmed  the patch, it seems now we're mixing the fsync(hsync) sequence id with sync(hflush), and if a sync call with higher sequenceId got executed before a fsync call (due to multiple syncRunners), the fsync call is actually written in a hflush way rather than expected. This issue was proposed and discussed back in HBASE-5954 around [this comment|https://issues.apache.org/jira/browse/HBASE-5954?focusedCommentId=14222296&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14222296], so are we ignoring this intentionally or by mistake? Worth a double check? Thanks. [~stack] [~apurtell] [~anoop.hbase]

> Configurable default durability for synchronous WAL
> ---------------------------------------------------
>
>                 Key: HBASE-19024
>                 URL: https://issues.apache.org/jira/browse/HBASE-19024
>             Project: HBase
>          Issue Type: Improvement
>          Components: wal
>            Reporter: Vikas Vishwakarma
>            Assignee: Harshal Jain
>            Priority: Critical
>             Fix For: 3.0.0, 2.1.0, 1.5.0
>
>         Attachments: HBASE-19024-master.v10.patch, HBASE-19024.branch-1.2.001.patch, HBASE-19024.branch-1.2.002.patch, HBASE-19024.branch-1.2.003.patch, HBASE-19024.branch-1.2.004.patch, HBASE-19024.branch-1.2.005.patch, branch-1.branch-1.patch, branch-1.v1.branch-1.patch, master.patch, master.v2.patch, master.v3.patch, master.v5.patch, master.v5.patch, master.v6.patch, master.v9.patch
>
>
> At present we do not have an option to hsync WAL edits to the disk for better durability. In our local tests we see 10-15% latency impact of using hsync instead of hflush which is not very high.  
> We should have a configurable option to hysnc WAL edits instead of just sync/hflush which will call the corresponding API on the hadoop side. Currently HBase handles both SYNC_WAL and FSYNC_WAL as the same calling FSDataOutputStream sync/hflush on the hadoop side. This can be modified to let FSYNC_WAL call hsync on the hadoop side instead of sync/hflush. We can keep the default value to sync as the current behavior and hsync can be enabled based on explicit configuration.



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