[
https://issues.apache.org/jira/browse/HDFS-8164?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14942017#comment-14942017
]
Yongjun Zhang commented on HDFS-8164:
-------------------------------------
Hi [~xiaochen],
Thanks for the revision. Some more comments about
{code}
1523 long getCTime() {
1524 return fsImage.getStorage().getCTime();
1525 }
{code}
1. As [~cnauroth] described, the time is
* initialized when NN is formated
* updated upon an upgrade
Ideally we can have two time: FormatTime, UpgadeTime to be clear what it means.
Lack of two names, I think we can continue to use getCTime, but we should have
javadoc for {{ long getCTime()}} to state what this time meant.
2. {{fsimage}} can be null, and {{getSotarge()}} can return null. Suggest to
check null before access. If we have one API at each layer, then only need to
check one there.
Thanks.
> cTime is 0 in VERSION file for newly formatted NameNode.
> --------------------------------------------------------
>
> Key: HDFS-8164
> URL: https://issues.apache.org/jira/browse/HDFS-8164
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: namenode
> Affects Versions: 2.0.3-alpha
> Reporter: Chris Nauroth
> Assignee: Xiao Chen
> Priority: Minor
> Attachments: HDFS-8164.001.patch, HDFS-8164.002.patch,
> HDFS-8164.003.patch
>
>
> After formatting a NameNode and inspecting its VERSION file, the cTime
> property shows 0. The value does get updated to current time during an
> upgrade, but I believe this is intended to be the creation time of the
> cluster, and therefore the initial value of 0 before an upgrade can cause
> confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
