[ 
https://issues.apache.org/jira/browse/HBASE-17085?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15663856#comment-15663856
 ] 

ramkrishna.s.vasudevan commented on HBASE-17085:
------------------------------------------------

'highestProcessedAppendTxid' increases. But my point was that when the code 
comes to appendAndSync when we try to process one highestProcessedAppendTxid at 
that time we have almost 40 to 50 sync future that is greater than the current 
'highestProcessedAppendTxid'. 
so again when the next appendAndSync happens we advance to the next 
'highestProcessedAppendTxid' say from 300 to 304 but the syncFuture is now with 
375 . Just for eg am saying.
So we have lot of syncFuture waiting to get to the done state.
Am away mostly for the coming days. Next two days can be online for sometime 
after that on a vacation.


> AsyncFSWAL may issue unnecessary AsyncDFSOutput.sync
> ----------------------------------------------------
>
>                 Key: HBASE-17085
>                 URL: https://issues.apache.org/jira/browse/HBASE-17085
>             Project: HBase
>          Issue Type: Sub-task
>          Components: wal
>    Affects Versions: 2.0.0
>            Reporter: Duo Zhang
>            Assignee: Duo Zhang
>             Fix For: 2.0.0
>
>         Attachments: HBASE-17085-v1.patch, HBASE-17085.patch
>
>
> The problem is in appendAndSync method, we will issue an  AsyncDFSOutput.sync 
> if syncFutures is not empty. The SyncFutures in syncFutures can only be 
> removed after an AsyncDFSOutput.sync comes back, so before the 
> AsyncDFSOutput.sync actually returns, we will always issue an  
> AsyncDFSOutput.sync after an append even if there is no new sync request.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to