Misty Stanley-Jones has posted comments on this change.

Change subject: No default partitioning blog post

Patch Set 2:


File _posts/2016-06-02-no-default-partitioning.md:

Line 15: partitioning behavior when creating new tables. The Kudu client will 
create a
If you don't specify a partitioning configuration, the Kudu client prior to 0.9 
creates a table with a _single tablet_. This is almost always a bad 
configuration, because you can't take advantage of any parallelism during 
writes or scans. We configured Kudu this way because we didn't anticipate that 
people would be so excited to get started that they would miss this step.

Line 23: the primary key is a better default, however even this is not a robust
s/however... but this approach has its own draw-backs.

Line 24: all-around default. Hash-partitioning can have non-obvious performance

Line 25: implications when scanning with predicates over primary key columns, 
and it's
Do we need an example of 'scanning with predicates over primary key columns'?

Line 28: Given that there is no bullet-proof default, and that changing the 
s/Given that/Since

s/and that changing/and changing

Line 29: configuration after table creation is impossible, and changing the 
can you rewrite this so that the last part of the sentence is separate? The 
backward-compatible part.

Line 30: partitioning strategy is a backwards incompatible change, we have 
decided to
Maybe link 'we have decided' to the Gerrit review where the discussion happened

Line 32: specify a partitioning configuration, or table creation will fail. It 
is still
s/It is still possible/You can

Line 38: configuration should be quite straightforward. The examples below add 
s/quite straightforward/simple

Line 39: partitioning, but it's also possible to specify range partitioning or a
s/it's also possible/you can also

Line 43: 
== C++ Client

(what I mean is make headings for C++ Client and Java Client)

Line 72: 
Maybe make Impala a sibling heading with C++ Client and Java Client

Line 86
Do we want to include anything about how to check your parallelism? Maybe point 
to one of the charts in CM? Or is there something in the metricz endpoint?

To view, visit http://gerrit.cloudera.org:8080/3277
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I14ace416b3eb5b61bb522a388e564be2e7dce3a9
Gerrit-PatchSet: 2
Gerrit-Project: kudu
Gerrit-Branch: gh-pages
Gerrit-Owner: Dan Burkert <d...@cloudera.com>
Gerrit-Reviewer: Adar Dembo <a...@cloudera.com>
Gerrit-Reviewer: Dan Burkert <d...@cloudera.com>
Gerrit-Reviewer: Jean-Daniel Cryans
Gerrit-Reviewer: Misty Stanley-Jones <mi...@apache.org>
Gerrit-Reviewer: Todd Lipcon <t...@apache.org>
Gerrit-HasComments: Yes

Reply via email to