Mike Percy has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/9069 )

Change subject: KUDU-2126: Add conditional check to prevent unnecessary fsyncs
......................................................................


Patch Set 2:

(9 comments)

http://gerrit.cloudera.org:8080/#/c/9069/2//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/9069/2//COMMIT_MSG@9
PS2, Line 9:
nit: a unit test


http://gerrit.cloudera.org:8080/#/c/9069/2//COMMIT_MSG@11
PS2, Line 11: add
nit: adds


http://gerrit.cloudera.org:8080/#/c/9069/2//COMMIT_MSG@12
PS2, Line 12: another
nit: additional


http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/integration-tests/delete_tablet-itest.cc
File src/kudu/integration-tests/delete_tablet-itest.cc:

http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/integration-tests/delete_tablet-itest.cc@169
PS2, Line 169: KUDU-2126 : Ensure that DeleteTablet RPCs does not do double 
fsyncs.
how about:

  // Regression test for KUDU-2126: Ensure that a DeleteTablet() RPC on a 
tombstoned tablet does not cause any fsyncs.


http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/integration-tests/delete_tablet-itest.cc@191
PS2, Line 191:   ASSERT_EVENTUALLY([&] {
Is this test flaky if you remove this ASSERT_EVENTUALLY? I don't think this is 
required because workload.Setup() should not return until the tablet has been 
created.


http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tablet/tablet_metadata.h
File src/kudu/tablet/tablet_metadata.h:

http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tablet/tablet_metadata.h@379
PS2, Line 379:   int flush_count_for_tests_;
This seems fine, but I wonder if you considered creating a metric for # of 
fsyncs.


http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tablet/tablet_metadata.cc
File src/kudu/tablet/tablet_metadata.cc:

http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tablet/tablet_metadata.cc@554
PS2, Line 554:   flush_count_for_tests_++;
I think this line should go inside TabletMetadata::ReplaceSuperBlockUnlocked() 
since that is where the actual flush is taking place. The flush occurs during 
the call to pb_util::WritePBContainerToPath().


http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tserver/ts_tablet_manager.cc
File src/kudu/tserver/ts_tablet_manager.cc:

http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tserver/ts_tablet_manager.cc@797
PS2, Line 797: KUDU-2126 : i
No need to mention KUDU-2126 here. It should be mentioned in the regression 
test, instead.


http://gerrit.cloudera.org:8080/#/c/9069/2/src/kudu/tserver/ts_tablet_manager.cc@798
PS2, Line 798: p
style nit: end comments with punctuation (i.e. a period). We follow the Google 
C++ style guide and this guideline can be found at 
https://google.github.io/styleguide/cppguide.html#Punctuation,_Spelling_and_Grammar



--
To view, visit http://gerrit.cloudera.org:8080/9069
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Idde8b413374f43ca2ef09339f0f412208f03685e
Gerrit-Change-Number: 9069
Gerrit-PatchSet: 2
Gerrit-Owner: Jeffrey F. Lukman <[email protected]>
Gerrit-Reviewer: Jeffrey F. Lukman <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <[email protected]>
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-Comment-Date: Sat, 20 Jan 2018 01:57:33 +0000
Gerrit-HasComments: Yes

Reply via email to