[
https://issues.apache.org/jira/browse/HDFS-15990?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17325930#comment-17325930
]
Stephen O'Donnell commented on HDFS-15990:
------------------------------------------
Both strategies are different ways of loading in parallel. I did not look into
HDFS-7784, I am not sure if I have seen it before you mentioned it here.
My intention was that HDFS-14617 would only be on the 3.x branch, but other
people wanted it backported to branch-2. Rollback should not be a problem in
practice, as the feature is off by default. Once you are running a version that
supports the feature AND are sure you are not going to roll back, you can
enable it and then ensure any further upgrades also have the feature in place.
> No need to write to sub_section when serialize SnapshotDiff Section
> -------------------------------------------------------------------
>
> Key: HDFS-15990
> URL: https://issues.apache.org/jira/browse/HDFS-15990
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: hdfs
> Affects Versions: 2.10.0
> Reporter: Xuze Yang
> Priority: Minor
>
> In FSImageFormatPBSnapshot.serializeSnapshotDiffSection(), the following code
> exists:
> {code:java}
> if (i % parent.getInodesPerSubSection() == 0) {
> parent.commitSubSection(headers,
> FSImageFormatProtobuf.SectionName.SNAPSHOT_DIFF_SUB);
> }{code}
> It aims to serialize SnapshotDiff information into several sub_sections(i.e.
> additional sub_sections information will be written to FileSummary Section).
> But in FSImageFormatPBSnapshot.loadSnapshotDiffSection(), it treats
> SnapshotDiffSection as a whole, rather than several sub_sections. So it's no
> need to introduce sub_sections here.
>
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
