Todd Lipcon has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/8831 )

Change subject: KUDU-2237: Allow idle server connection scanning to be disabled
......................................................................


Patch Set 5:

(4 comments)

http://gerrit.cloudera.org:8080/#/c/8831/5//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/8831/5//COMMIT_MSG@20
PS5, Line 20: set to 0. This avoids the unnecessary overhead of scanning
I think -1 would make more sense as a special value. '0' sounds like 'keepalive 
disabled' which would mean that you would immediate close the connection any 
time there are no pending calls, no?


http://gerrit.cloudera.org:8080/#/c/8831/5/src/kudu/rpc/rpc-test.cc
File src/kudu/rpc/rpc-test.cc:

http://gerrit.cloudera.org:8080/#/c/8831/5/src/kudu/rpc/rpc-test.cc@402
PS5, Line 402:     continue;
> warning: redundant continue statement at the end of loop statement [readabi
agreed with tidybot here


http://gerrit.cloudera.org:8080/#/c/8831/5/src/kudu/rpc/rpc-test.cc@407
PS5, Line 407: TestConnectionNoTimeout
'AlwaysKeepalive" is more accurate than 'NoTimeout' since we use the term 
'timeout' to refer to call timeouts, not connection destruction


http://gerrit.cloudera.org:8080/#/c/8831/5/src/kudu/rpc/rpc-test.cc@435
PS5, Line 435:     // SlowCallback() will spin until half of the calls have 
been made. Sleeps between calls
             :     // to cause the single reactor thread spin for extended 
period of time. With a low enough
             :     // 'keepalive_time_ms_', the server connection will go idle 
and get disconnected while
             :     // there are pending calls in the outbound queue in the 
client messenger, triggering
             :     // KUDU-279. Setting 'keepalive_time_ms_' to 0 should 
prevent this from happening.
I'm not entirely following this test. You're trying to trigger KUDU-279, but 
then you say it won't happen? Is this test trying to show that, in the presence 
of KUDU-279, the 0-keepalive acts as a workaround to that? It seems odd to use 
the presence of one bug to form a unit test for a separate feature, vs a much 
more simple test which just makes a single call, sleeps a few seconds, and 
makes sure the connection is still established.



--
To view, visit http://gerrit.cloudera.org:8080/8831
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I6161b9e753f05620784565a417d248acf8e7050a
Gerrit-Change-Number: 8831
Gerrit-PatchSet: 5
Gerrit-Owner: Michael Ho <k...@cloudera.com>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon <t...@apache.org>
Gerrit-Comment-Date: Tue, 19 Dec 2017 03:26:54 +0000
Gerrit-HasComments: Yes

Reply via email to