[
https://issues.apache.org/jira/browse/GORA-144?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14027075#comment-14027075
]
Renato Javier Marroquín Mogrovejo commented on GORA-144:
--------------------------------------------------------
Hi guys,
I also have been thinking about, and there are some decisions we should make on
this one. We are trying to support different hadoop versions but because of the
data stores we support, not because our code base needs it directly (or at
least it shouldn't IMHO). Some other projects use profiles because their code
base needs them, Gora could be run inside MapReduce but that doesn't mean we
should depend on it (it's a totally different story with Avro).
I think the two options we have are:
1) Using profiles + conditional execution -> This is fine, and has been done in
other projects, but I think it leads to more complexity than we need. I mean we
will have to live with this complexity.
2) Using different modules for each specific data store version. We would have
to create something like gora-hbase-core which defines the API and what is
needed, and separate versions for the hadoop versions we want to support. Then
if people want to enhance specific versions they could do it separately without
messing with gora-hbase-core (which would contain all the serialization
mechanism which doesn't depend on HBase itself), and we could support legacy
code.
> Upgrade gora so to be compatible with hadoop 2
> ----------------------------------------------
>
> Key: GORA-144
> URL: https://issues.apache.org/jira/browse/GORA-144
> Project: Apache Gora
> Issue Type: Task
> Components: avro, build process, maven
> Affects Versions: 0.2
> Reporter: raf shin
> Assignee: Talat UYARER
> Priority: Minor
> Fix For: 0.5
>
> Attachments: GORA-144-v1.patch, GORA-144-v2.patch, GORA-144v3.patch
>
>
> TaskAttemptContext and JobContext which were a class in previous hadoop
> versions are now interfaces.
> Compiling gora with new hadoop version 2 fails.
> BinaryDecoder and BinaryEncoder in Avro 1.5.4 are also now interfaces which
> were previously class.
> avro.ipc.ByteBufferOutputStream and avro.ipc.ByteBufferInputStream are also
> moved to avro.util package.
> There might be much more incompatibilities there which we have not seen yet.
> We gave up upgrading our nutchgora to the new hadoop 2 version.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
