Andrew Wong has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/11639 )

Change subject: [compaction] KUDU-2056: Expose a metric for how much a tablet 
needs to be compacted
......................................................................


Patch Set 4:

> Patch Set 4:
>
> (1 comment)
>
> > We discussed this offline: currently the placement of the computation of 
> > the rowset height is such that it is performed every time we swap in new 
> > rowsets. Conceptually this makes sense, but a quirk that exists is that 
> > we'll swap rowsets twice during merges/flushes, rendering one of these 
> > computations unnecessary. Will's added a TODO to cache the height to avoid 
> > this.
>
> Yes, I'm also curious about what kind of overhead this adds. Do we only swap 
> rowsets during an IO operation (i.e. flush or compaction)? Or at other times 
> too?

Right, AFAICT AtomicSwapRowSets*() is always accompanied by some sort of flush.


--
To view, visit http://gerrit.cloudera.org:8080/11639
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I98493b901d37bb278167ba2fe98d322a86a1f0f9
Gerrit-Change-Number: 11639
Gerrit-PatchSet: 4
Gerrit-Owner: Will Berkeley <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Andrew Wong <[email protected]>
Gerrit-Reviewer: Attila Bukor <[email protected]>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Mike Percy <[email protected]>
Gerrit-Reviewer: Todd Lipcon <[email protected]>
Gerrit-Reviewer: Will Berkeley <[email protected]>
Gerrit-Comment-Date: Wed, 17 Oct 2018 18:02:34 +0000
Gerrit-HasComments: No

Reply via email to