[
https://issues.apache.org/jira/browse/HBASE-15265?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15154485#comment-15154485
]
stack commented on HBASE-15265:
-------------------------------
If creation is fail-fast, is presumption that we go try create again... and
keep at it till we succeed? When would we give up?
I think abort if can't roll the log is dumb but basic and easy to understand.
Do current WAL providers just rely on DFSClient timeouts? That is why they are
more 'dogged' [1]? If this latter, then it would make sense if asyncwal does a
bunch of retrying... before throwing a fail up to logroller.
bq. The shutdown timing is a little strange.
The shutdown sequence come about by trial and error. You are changing the
behavior of one of the main components. The shutdown sequence may have to
change. The way you describe it, current WAL implementation could get into same
state (only for the sync timeout).
1
https://www.google.com/search?q=define%3Adogged&oq=define%3Adogged&aqs=chrome..69i57j69i58.4223j0j7&sourceid=chrome&es_sm=91&ie=UTF-8
> Implement an asynchronous FSHLog
> --------------------------------
>
> Key: HBASE-15265
> URL: https://issues.apache.org/jira/browse/HBASE-15265
> Project: HBase
> Issue Type: Sub-task
> Reporter: Duo Zhang
> Assignee: Duo Zhang
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
