impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael Ho (Code Review)" <ger...@cloudera.org>
Subject [Impala-ASF-CR] IMPALA-5744: Add dummy 'use krpc' flag and create DataStream interface
Date Wed, 02 Aug 2017 21:59:21 GMT
Michael Ho has posted comments on this change.

Change subject: IMPALA-5744: Add dummy 'use_krpc' flag and create DataStream interface
......................................................................


Patch Set 5:

(10 comments)

http://gerrit.cloudera.org:8080/#/c/7542/4/be/src/runtime/data-stream-mgr-base.h
File be/src/runtime/data-stream-mgr-base.h:

Line 54:   virtual Status CloseSender(const TUniqueId& fragment_instance_id, PlanNodeId
dest_node_id,
nit: long line


http://gerrit.cloudera.org:8080/#/c/7542/5/be/src/runtime/data-stream-mgr-base.h
File be/src/runtime/data-stream-mgr-base.h:

Line 35: /// node.
This is a pure virtual class for the thrift and KRPC implementation. Please also add a TODO
to remove this class and rename all clients once the parallel implementation is removed.


Line 50:                  const TRowBatch& thrift_batch, int sender_id) = 0;
nit: wrong indent


PS5, Line 49:   virtual Status AddData(const TUniqueId& fragment_instance_id, PlanNodeId
dest_node_id,
            :                  const TRowBatch& thrift_batch, int sender_id) = 0;
As discussed offline, since this interface will diverge across the two implementations, we
can do either of the following:

1. remove AddData() from the virtual base class and make the client do an explicit type cast
to the underlying type.

2. have a common interface and hide the arguments for the two different implementations inside
a union.

Please consider the options above. I personally think (1) will be least disruptive to the
existing code and the eventual removal of the parallel implementation.


Line 54:   virtual Status CloseSender(const TUniqueId& fragment_instance_id, PlanNodeId
dest_node_id,
nit: long line


http://gerrit.cloudera.org:8080/#/c/7542/5/be/src/runtime/data-stream-recvr-base.h
File be/src/runtime/data-stream-recvr-base.h:

Line 28: /// Single receiver of an m:n data stream.
Please add a more meaningful description of this base class.


PS5, Line 39: the
nit: long line


http://gerrit.cloudera.org:8080/#/c/7542/4/be/src/runtime/exec-env.cc
File be/src/runtime/exec-env.cc:

PS4, Line 167: errorwith
nit: missing space


http://gerrit.cloudera.org:8080/#/c/7542/5/be/src/runtime/exec-env.cc
File be/src/runtime/exec-env.cc:

PS5, Line 167:    // TODO: Replace fatal errorwith KRPCDataStreamMgr implementation when it
is
             :     // supported.
             :     ABORT_WITH_ERROR("The 'use_krpc' flag is not supported yet. "
             :         "Please disable and restart cluster");
Please consider adding a stub implementation of DataStreamMgr for KRPC instead of warning
here.


http://gerrit.cloudera.org:8080/#/c/7542/5/be/src/service/impala-server.cc
File be/src/service/impala-server.cc:

PS5, Line 1921:   // TODO: Remove DCHECK when KRPC is supported.
              :   DCHECK(!FLAGS_use_krpc);
How about we add a stub implementation for KRPC and keep the DCHECK in the stub implementation
instead.


-- 
To view, visit http://gerrit.cloudera.org:8080/7542
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I5d52245154e910529a68f53049520238eca16241
Gerrit-PatchSet: 5
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Sailesh Mukil <sailesh@cloudera.com>
Gerrit-Reviewer: Henry Robinson <henry@cloudera.com>
Gerrit-Reviewer: Michael Ho <kwho@cloudera.com>
Gerrit-Reviewer: Sailesh Mukil <sailesh@cloudera.com>
Gerrit-HasComments: Yes

Mime
View raw message