David Ribeiro Alves has posted comments on this change.

Change subject: KUDU-798 (part 4) Add a TimeManager to manage safe time 
advancement
......................................................................


Patch Set 14:

(12 comments)

http://gerrit.cloudera.org:8080/#/c/5300/14/src/kudu/consensus/time_manager-test.cc
File src/kudu/consensus/time_manager-test.cc:

PS14, Line 52:   // Waits (forever) for the TimeManager's safe time to have 
advanced to 'safe_time'.
             :   // Returns a latch that allows to wait for TimeManager to 
consider 'safe_time' safe.
             :  
> the two lines of this comment seem contradictory with each other. The secon
Done


Line 58:       time_manager_->WaitUntilSafe(safe_time, MonoTime::Max());
> should this CHECK_OK?
Done


Line 94:   ASSERT_OK(time_manager_->MessageReceivedFromLeader
> weird wrapping
Done


Line 101:   MonoTime after_small = MonoTime::Now();
> I tihnk with recent changes we can just do:
Done


http://gerrit.cloudera.org:8080/#/c/5300/14/src/kudu/consensus/time_manager.cc
File src/kudu/consensus/time_manager.cc:

Line 107:     Timestamp t(message.timestamp());
> unused?
Done


Line 170: void TimeManager::AdvanceSafeTimeAndWakeUpWaiters(Timestamp 
safe_time) {
> should this be named 'Unlocked'?
Done


PS14, Line 190: (
> also unlocked?
Done


http://gerrit.cloudera.org:8080/#/c/5300/14/src/kudu/consensus/time_manager.h
File src/kudu/consensus/time_manager.h:

PS14, Line 40: Because of this the read will be repeatable
> this probably needs to say something like "If the scanner additionally uses
Done


PS14, Line 52: sintony
> typo?
this is a "portuguesism" :P I looked in the dictionary and it seemed to exist 
:). Done


Line 70:   TimeManager(scoped_refptr<server::Clock> clock,
> nit: no need for wrap
Done


PS14, Line 95: CHECK failure if it isn't)
> curious why the inconsistency that here it's a CHECK failure but above it's
because of races in changes of leadership status. consensus might not be leader 
anymore by the time prepare tries to assign a timestamp so the above shouldn't 
crash. This one is called on UpdateConsensus() though, so the timemanager 
should not be in leader mode.


PS14, Line 107:  'message's timestamp
> 'safe_time'
Done


-- 
To view, visit http://gerrit.cloudera.org:8080/5300
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I0bb2b1d2590ed7ead6f1980f9572be10444bb81b
Gerrit-PatchSet: 14
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: David Ribeiro Alves <dral...@apache.org>
Gerrit-Reviewer: Alexey Serbin <aser...@cloudera.com>
Gerrit-Reviewer: David Ribeiro Alves <dral...@apache.org>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <mpe...@apache.org>
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon <t...@apache.org>
Gerrit-HasComments: Yes

Reply via email to