[
https://issues.apache.org/jira/browse/HDFS-1580?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13008886#comment-13008886
]
Todd Lipcon commented on HDFS-1580:
-----------------------------------
Hi Ivan,
Trying to integrate this with my HDFS-1073 work, running into some things that
feel wrong about the JournalManager API...
It seems strange that every edit makes a call to getOutputStream(txid) rather
than simply calling it once at each roll point.
I think it makes the most sense to introduce the idea of a "log segment" - in
the file case it corresponds to a single edit file. In the BK case it might be
a ledger. In other cases it might be a no-op.
The state machine for logging then becomes something like:
- open
- beginLogSegment(txid 1): returns an EditLogOutputStream
- write(...) 5 times
- roll:
-- endLogSegment(1, 6)
-- startLogSegment(7)
- write more data, etc
Does this seem to make sense to you?
> 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
