Alexey Serbin has posted comments on this change.

Change subject: [client-test] one more test for AUTO_FLUSH_BACKGROUND

Patch Set 4:

> my point was that we should make sure to stress it a bit so that
 > over time we will hit flush/flush async over different stages of
 > the auto flushing.  The way you're doing it now, you always call
 > flush at precise intervals, my suggestion was that you spin insert
 > enough rows to actually get background batches being inserted while
 > calling flush/flush async at random times, possibly increasing the
 > coverage. would that make sense?

Getting it stressed as you described certainly makes sense.  However, given 
current run times of the test I would not put that test in here.  Probably, we 
can add such a test into the integration test bundle?  Or it's OK to add a test 
which runs additional 5 minutes in TSAN configuration?

As I see it, the purpose of this test is to make sure there aren't errors, 
deadlocks or missed/unflushed batchers while calling Flush/FlushAsync in 
AUTO_FLUSH_BACKGROUND mode.  From that perspective, exploring behavior of the 
client/server in all phases besides the very first one it is no different than 
calling Flush/AsyncFlush in any other regular session.  The only phase of 
interest in that context is when there is contention on detaching the current 
batcher by auto-flusher and the call of explicit flush.

What do you think?

To view, visit
To unsubscribe, visit

Gerrit-MessageType: comment
Gerrit-Change-Id: I3bd5d248d4d44393689c8da81ed669395c393257
Gerrit-PatchSet: 4
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Alexey Serbin <>
Gerrit-Reviewer: Alexey Serbin <>
Gerrit-Reviewer: David Ribeiro Alves <>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-HasComments: No

Reply via email to