Define Table Partitions Doc Update - "partition by index"

[andreimatei]

Andrei Matei (andreimatei) commented:

Re: Define Table Partitions

The Partition using secondary index section is pretty unclear. The title suggests that it's about partitioning secondary indexes, but it mostly talks about creating a non-partitioned index.
Separately, when describing the use cases for the non-partitioned index, it conspicuously fails to mention that, by creating a non-partitioned index, you might reduce or eliminate the benefits of partitioning anything in that table (since, if you're using partitioning for multi-region, writing to this index will be a cross-region operation, and joins between this index and the PK will also be a multi-region operation).

Generally, the page as a whole is not very clear in explaining the differences between partitioning the PK and partitioning another index, the fact that each index is partitioned individually (at no point does the page give an example of partitioning a secondary index), and the fact that when you use the partitioning syntax on a create table statement, you're partitioning only the primary key (I think so, at least).

Other suggestions for improving the page gathered at the same time: #9205

cc @nvanbenschoten

Jira Issue: DOC-873

[lnhsingh]

@ericharmeling / @ianjevans - not sure if this falls under you two or maybe @rmloveland. Feel free to reassign!