Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/8393 )

Change subject: KUDU-2200: provide better diagnostics when connecting to a 
subset of masters
......................................................................


Patch Set 2:

(4 comments)

http://gerrit.cloudera.org:8080/#/c/8393/1/java/kudu-client/src/test/java/org/apache/kudu/client/TestConnectToCluster.java
File 
java/kudu-client/src/test/java/org/apache/kudu/client/TestConnectToCluster.java:

http://gerrit.cloudera.org:8080/#/c/8393/1/java/kudu-client/src/test/java/org/apache/kudu/client/TestConnectToCluster.java@108
PS1, Line 108:     // One of the connections should have succeeded.
             :     assertEquals(1, successes);
> Yah I'm just being kind of conservative.  If the client always hard fails w
I see.  For some reason I thought this patch was to introduce a stricter policy 
for specifying master end-points, but it seems it's more about proper error 
reporting and actionable message in that case.  I think it makes sense as well. 
 Thanks for clarifying.


http://gerrit.cloudera.org:8080/#/c/8393/2/src/kudu/integration-tests/master_replication-itest.cc
File src/kudu/integration-tests/master_replication-itest.cc:

http://gerrit.cloudera.org:8080/#/c/8393/2/src/kudu/integration-tests/master_replication-itest.cc@308
PS2, Line 308: LOG(ERROR) << resp.DebugString();
nit: is it crucial to output this, especially as an error?


http://gerrit.cloudera.org:8080/#/c/8393/2/src/kudu/integration-tests/master_replication-itest.cc@333
PS2, Line 333:   // - 0, in the case that no master had elected itself yet
             :   // - 1, in the case that one master had become leader by the 
time we connected.
Is it possible that leader re-election happens among masters just between 
clients are connected and successes becomes 2?  Most likely, it's not the case 
for 'fast' builds, but a TSAN build might be the case.


http://gerrit.cloudera.org:8080/#/c/8393/2/src/kudu/integration-tests/master_replication-itest.cc@335
PS2, Line 335: LE
It seems it should be EXPECT_GE().

Maybe, it's worth adding the same relaxed GE assert into the Java test as well? 
 There I saw strict EQ assert.



--
To view, visit http://gerrit.cloudera.org:8080/8393
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I52f903e1aa5ae6948ca1ba6d4d856c3c9dc73d56
Gerrit-Change-Number: 8393
Gerrit-PatchSet: 2
Gerrit-Owner: Todd Lipcon <t...@apache.org>
Gerrit-Reviewer: Alexey Serbin <aser...@cloudera.com>
Gerrit-Reviewer: Dan Burkert <danburk...@apache.org>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon <t...@apache.org>
Gerrit-Comment-Date: Tue, 31 Oct 2017 06:19:25 +0000
Gerrit-HasComments: Yes

Reply via email to