[
https://issues.apache.org/jira/browse/HDFS-1580?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13009138#comment-13009138
]
Ivan Kelly commented on HDFS-1580:
----------------------------------
This depends on whether we want the roll operation exposed to the user of the
JournalManager. In the past this was required as roll had to happen to allow a
checkpoint to take place. However, once HDFS-1580 goes in, its not strictly
necessary because rolling is stateless. The JournalManager could roll in the
background without the user of the class having any knowledge of the roll. To
the user, the JournalManager is simply something to open, write to and close.
Do you think this simplification is worthwhile?
I agree that getOutputStream(txid) everywhere is a bit messy. It also doesnt
work. If a roll happens while the OutputStream has been given out the editlog
may end up writing to two different streams which could mess up transaction
ids. This isn't in any way unsolvable, but I need to think about it more.
> Add interface for generic Write Ahead Logging mechanisms
> --------------------------------------------------------
>
> Key: HDFS-1580
> URL: https://issues.apache.org/jira/browse/HDFS-1580
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Ivan Kelly
> Attachments: HDFS-1580+1521.diff, HDFS-1580.diff,
> generic_wal_iface.pdf, generic_wal_iface.pdf, generic_wal_iface.txt
>
>
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
