[
https://issues.apache.org/jira/browse/HADOOP-6904?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12966768#action_12966768
]
Sanjay Radia commented on HADOOP-6904:
--------------------------------------
> .. But adding or removing a method would not require a change of this [major
> #] value.
In the original Major-minor proposal a deletion of a method does require a
change in Major#.
* Use case 1. Remove method foo() because we think it is not needed and we
are willing to break client apps.
Old clients connect and realize that the Major# has changed and disconnect. Old
applications cannot run against new server. We detect this, not when foo() is
called, but whent
getProxy() is called. If the server has broken compatibility I want clean
failure rather then a partial failure when the missing method is called.
* Use case 2. Service has a method bar and we add method fasterBar(). getProxy
succeeds because it is a compatible change.
New client side checks, at the time of the method call, to see if fasterBar
is supported. If not then calls bar instead.
For both use cases there is a mismatch of the list of methods. But in case 2
it is okay (bar is fall-back for fasterBar).
In use case 1 it is not okay. I guess one could argue that we should let the
app run and if it uses the removed method then, and only then, fail. I am
uncomfortable about that but and willing to be convinced. The point is that the
two approaches are NOT equivalent.
My current thought is that Major number changes when you delete a method,
change the signature of the method, of the serialization of the method.
> A baby step towards inter-version communications between dfs client and
> NameNode
> --------------------------------------------------------------------------------
>
> Key: HADOOP-6904
> URL: https://issues.apache.org/jira/browse/HADOOP-6904
> Project: Hadoop Common
> Issue Type: New Feature
> Components: ipc
> Affects Versions: 0.22.0
> Reporter: Hairong Kuang
> Assignee: Hairong Kuang
> Fix For: 0.22.0
>
> Attachments: majorMinorVersion.patch, majorMinorVersion1.patch,
> rpcVersion.patch, rpcVersion1.patch
>
>
> Currently RPC communications in Hadoop is very strict. If a client has a
> different version from that of the server, a VersionMismatched exception is
> thrown and the client can not connect to the server. This force us to update
> both client and server all at once if a RPC protocol is changed. But sometime
> different versions do not mean the client & server are not compatible. It
> would be nice if we could relax this restriction and allows us to support
> inter-version communications.
> My idea is that DfsClient catches VersionMismatched exception when it
> connects to NameNode. It then checks if the client & the server is
> compatible. If yes, it sets the NameNode version in the dfs client and allows
> the client to continue talking to NameNode. Otherwise, rethrow the
> VersionMismatch exception.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
