[
https://issues.apache.org/jira/browse/HBASE-2580?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12869818#action_12869818
]
stack commented on HBASE-2580:
------------------------------
Only blog should know how file name is formatted so move the parse of filename
there. The encoding should also be in blog. What if u are passed a prefix
string that is illegal as part of a filename in hdfs? The javadoc on prefix
needs to be beefier suggesting hostnAme and port as prefix when in distributed
en ironmenr and thAt blog will be used if none supplied.
Fix Above and I am +1 commit
> Make the hlog file names unique
> -------------------------------
>
> Key: HBASE-2580
> URL: https://issues.apache.org/jira/browse/HBASE-2580
> Project: Hadoop HBase
> Issue Type: Improvement
> Reporter: Jean-Daniel Cryans
> Assignee: Jean-Daniel Cryans
> Fix For: 0.21.0
>
> Attachments: HBASE-2580-v2.patch, HBASE-2580.patch
>
>
> Currently the HLog archiving thread adds another timestamp to ensure the file
> names are "unique". This is ugly and makes it harder to track the HLogs
> movements from outside (like replication). Instead we could use a UUID as the
> file name.
> I was discussing with Stack the need of keeping a timestamp in the file
> name... I think we can't get rid of it since we need to read the HLogs in
> sequence when splitting.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
