[
https://issues.apache.org/jira/browse/HADOOP-5188?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12707535#action_12707535
]
Flavio Paiva Junqueira commented on HADOOP-5188:
------------------------------------------------
It is perhaps not clear what problem point number 3 of the concrete tasks is
trying to solve. If I understand it correctly, there is an opcode
(OP_JSPOOL_START) that the backup node uses to mark an edits stream and
determine the last synchronization point. I believe the idea is to leverage
this mechanism to close one ledger and create another ledger when using
BookKeeper. In general, this mechanism flags to any log device where
synchronization points are, and let each device handle accordingly. The
important property here is that all devices will observe the same
synchronization points and in the same positions.
One problem with the implementation of this marking mechanism is that it is a
little obscure in the code, so you need to know that it is there. What I
believe the design document is proposing is to add an interface method that
forces a developer implementing the adapter for some new log device to get a
error at compile time if it doesn't implement such a method. For the streams
that do not support such markers, the method call would be a simple no-op.
Also, an implementation of the method isOperationSupported() might be
unnecessary if we do have such an interface method.
Comments or clarifications?
> Modifications to enable multiple types of logging
> --------------------------------------------------
>
> Key: HADOOP-5188
> URL: https://issues.apache.org/jira/browse/HADOOP-5188
> Project: Hadoop Core
> Issue Type: Improvement
> Components: dfs
> Affects Versions: 0.19.0
> Reporter: Luca Telloli
> Attachments: HADOOP-5188.patch, HADOOP-5188.pdf
>
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
