[
https://issues.apache.org/jira/browse/HDFS-5066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13751001#comment-13751001
]
Binglin Chang commented on HDFS-5066:
-------------------------------------
bq. I guess here you mean we need some unique id for all the INodes so that we
can check whether we have visited them while traversing the graph?
Not only to check visit status, for a graph(json or gv), each node need a
unique label as node identifier. I checked getObjectString it uses
object.hashcode which is based on memory address, googled, and looks like we
can't assume it is unique(across different jvms). It is only int32, how can it
be unique in a 64bit environment with 100Gs of memory.
bq. we may need to define some upper limit for the maximum depth or maximum
number of inodes
You are right, I think maximum number of nodes make sense, since depth can't
control the number of memory if there are too many children in directory.
> Inode tree with snapshot information visualization
> ---------------------------------------------------
>
> Key: HDFS-5066
> URL: https://issues.apache.org/jira/browse/HDFS-5066
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Binglin Chang
> Assignee: Binglin Chang
> Priority: Minor
> Attachments: HDFS-5066.v1.patch, visnap.png
>
>
> It would be nice to be able to visualize snapshot information, in order to
> ease the understanding of related data structures. We can generate graph from
> in memory inode links.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
