[
https://issues.apache.org/jira/browse/PHOENIX-1673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14662412#comment-14662412
]
ASF GitHub Bot commented on PHOENIX-1673:
-----------------------------------------
Github user elilevine commented on the pull request:
https://github.com/apache/phoenix/pull/104#issuecomment-128830531
Good stuff, @JeffreyLyonsD2L. Added a few questions and comments to the PR.
There is one more issue that it not tackled here that we need to at least
discuss IMHO. Phoenix's metadata tables, such as SYSTEM.CATALOG lead with
TENANT_ID VARCHAR column (take a look in QueryConstants.java). This works well
in a world where all multi-tenant tables can only have VARCHAR tenant id
columns. Your changes introduce the possibility that different multi-tenant
tables will have different types of tenant id columns.
What's the right solution here? The right solution is probably changing
TENANT_ID column type to BINARY in system tables. But that will make upgrades
for existing clusters a pain. Leave them as VARCHAR?
> Allow tenant ID to be of any integral data type
> -----------------------------------------------
>
> Key: PHOENIX-1673
> URL: https://issues.apache.org/jira/browse/PHOENIX-1673
> Project: Phoenix
> Issue Type: Improvement
> Affects Versions: 4.3.0
> Reporter: Mark Tse
> Labels: Newbie, multi-tenant
> Fix For: 4.4.1
>
>
> When creating multi-tenant tables and views, the column that identifies the
> tenant (first primary key column) must be of type 'VARCHAR' or 'CHAR'.
> It should be possible to relax this restriction to use any integral data
> type. The tenant ID from the connection property can be converted based on
> the data type of the first primary key column.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
