[
https://issues.apache.org/jira/browse/BOOKKEEPER-572?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13622592#comment-13622592
]
Ivan Kelly commented on BOOKKEEPER-572:
---------------------------------------
New patch addresses the roll-logic issue. Unfortunately BOOKKEEPER-564's
checkpoint changes didn't address the issue, as checkpointing needed to be
completely overhauled as the order of ops changed.
I've gotten rid of LastLogMark and replaced it with JournalMark which is an
immutable object. The Journal holds the JournalMark. When the journal flushs,
it creates a new journal mark object and flushes it. When a bookie adds an
entry to the journal, it will first take the mark from the journal. When the
callback occurs, and the bookie adds the entry to the ledger storage, then the
mark is passed to the sync thread. Then, when the sync thread runs, it knows to
what point of the journal has been stored in the ledger storage and uses it
accordingly. The nice thing about this is that the journal knows very little
about the mark, and the ledger storage doesn't need to know anything. The work
is done by the bookie, and the sync thread, which are the entities interested
in marking.
There's a few more changes to be made, like moving the reading and writing of
the mark out of Journal.java, but this is refactoring, so I don't want to do it
in this patch.
> Make the journal a write ahead log
> ----------------------------------
>
> Key: BOOKKEEPER-572
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-572
> Project: Bookkeeper
> Issue Type: Bug
> Reporter: Ivan Kelly
> Assignee: Ivan Kelly
> Fix For: 4.3.0
>
> Attachments:
> 0001-BOOKKEEPER-572-Write-to-the-journal-before-writing-t.patch,
> 0003-BOOKKEEPER-572-Write-to-the-journal-before-writing-t.patch,
> 0003-BOOKKEEPER-572-Write-to-the-journal-before-writing-t.patch,
> BookieServer-2013-02-22.snapshot
>
>
> A bookie adds to the LedgerStorage before writing to the journal. This is the
> fundamental problem behind BOOKKEEPER-447 and blocks a nice solution to
> BOOKKEEPER-530. By writing to the memory state before the journal, we exposed
> ourselves to bugs if the bookie crashed before we wrote to the journal. The
> entry may exist in index, but not in the entrylog, a situation which cannot
> be distinguished from an I/O error. The comments on BOOKKEEPER-447 goes into
> more details.
--
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
