hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anu Engineer (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-11921) Ozone: KSM: Unable to put keys with zero length
Date Tue, 12 Sep 2017 16:23:00 GMT

    [ https://issues.apache.org/jira/browse/HDFS-11921?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16163194#comment-16163194

Anu Engineer commented on HDFS-11921:

bq. This is needed for OzoneFileSystem as well, where this API is needed for directories.
Directories will have keys with no data.
good to know. So you are saying that the behaviour specified by [~vagarychen] is all we need.
bq.Right now, the behaviour is that if a length of 0 is given, key allocation will succeed
but following writes will throw exception due to writing out of range.

I am going to resolve this JIRA as fixed.

> Ozone: KSM: Unable to put keys with zero length
> -----------------------------------------------
>                 Key: HDFS-11921
>                 URL: https://issues.apache.org/jira/browse/HDFS-11921
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: ozone
>    Affects Versions: HDFS-7240
>            Reporter: Anu Engineer
>            Assignee: Mukul Kumar Singh
>            Priority: Minor
>              Labels: OzonePostMerge
>         Attachments: HDFS-11921-HDFS-7240.001.patch
> As part of working on HDFS-11909, I was trying to put zero length keys. I found that
put key refuses to do that. Here is the call trace, 
> bq.	at ScmBlockLocationProtocolClientSideTranslatorPB.allocateBlock 
> we check if the block size is greater than 0, which makes sense since we should not call
into SCM to allocate a block of zero size.
> However these 2 calls are invoked for creating the key, so that metadata for key can
be created, we should probably take care of this behavior here.
> bq. ksm.KeyManagerImpl.allocateKey
> bq. ksm.KeySpaceManager.allocateKey(KeySpaceManager.java:428)
> Another way to fix this might be to just allocate a block with at least 1 byte always,
which might be easier than special casing code.
> [~vagarychen] Would you like to fix this in the next patch you are working on ? 

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org

View raw message