[
https://issues.apache.org/jira/browse/HDFS-8489?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhe Zhang updated HDFS-8489:
----------------------------
Attachment: HDFS-8489.02.patch
Updating the patch to remove redundant logic between {{BlockInfo}} and
{{BlockInfoContiguous}}.
The main difference between {{BlockInfoStriped}} and {{BlockInfoContiguous}} is
that in {{BIStriped#triplets}}, the first {{dataBlockNum}} slots are ordered
based on internal block indices. Therefore the first {{dataBlockNum}} slots
could have null, and we need an indices array to interpret the slots after
{{dataBlockNum}}. So only {{addStorage}}, {{removeStorage}}, and {{numNodes}}
should stay abstract in {{BlockInfo}} and be separately implemented.
[~jingzhao] We discussed similar ideas under HDFS-7285 JIRAs. Let me know if
the above makes sense to you.
> Subclass BlockInfo to represent contiguous blocks
> -------------------------------------------------
>
> Key: HDFS-8489
> URL: https://issues.apache.org/jira/browse/HDFS-8489
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: namenode
> Affects Versions: 2.7.0
> Reporter: Zhe Zhang
> Assignee: Zhe Zhang
> Attachments: HDFS-8489.00.patch, HDFS-8489.01.patch,
> HDFS-8489.02.patch
>
>
> As second step of the cleanup, we should make {{BlockInfo}} an abstract class
> and merge the subclass {{BlockInfoContiguous}} from HDFS-7285 into trunk. The
> patch should clearly separate where to use the abstract class versus the
> subclass.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
