You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Chris Nauroth (JIRA)" <ji...@apache.org> on 2015/12/04 19:02:11 UTC

[jira] [Commented] (HADOOP-12614) Add a generic .isOffline() method to filesystems to probe availability

    [ https://issues.apache.org/jira/browse/HADOOP-12614?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15041875#comment-15041875 ] 

Chris Nauroth commented on HADOOP-12614:
----------------------------------------

bq. ...reflection games to check whether HDFS is in safe mode or not...

Hi [~stevel@apache.org].  For my curiosity, would you mind giving a link to the Spark code that does this?

I'm wondering how useful this proposal would really be.  It seems inherently subject to a time of check/time of use race condition.  A caller could write this...

{code}
if (!fs.isOffline()) {
  FSDataOutputStream os = fs.create(...);
}
{code}

...and it still wouldn't guarantee that the application never hits a {{SafeModeException}}.

Maybe seeing that Spark code in context would help clarify it for me.

> Add a generic .isOffline() method to filesystems to probe availability
> ----------------------------------------------------------------------
>
>                 Key: HADOOP-12614
>                 URL: https://issues.apache.org/jira/browse/HADOOP-12614
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: fs
>    Affects Versions: 2.7.1
>            Reporter: Steve Loughran
>            Priority: Minor
>
> Looking at some of the spark `HistoryServer` code, they do reflection games to check whether HDFS is in safe mode or not, games which vary with version and could be at risk of failing with the client/server split (fortunately, it's all client-side). Nor do the checks apply to other filesystems, which could have their own online/offline state.
> I propose adding the new methods {{FileSystem.isOffline()}}, {{FileContext.isOffline()}}, to return true if an FS knows that it is offline. For HDFS: Safe mode. For other filesystems? Maybe network state, disk being r/W, etc. Their choice. The default would be false: an FS is not offline,
> obvously, {{!isOffline()}} doesn't guarantee the FS is fully functional; that's why I propose {{isOffline()}}; less dangerous than the opposite values of {{isLive()}} or {{isAvailable()}}, which may be making promises which cannot hold 



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