[
https://issues.apache.org/jira/browse/KUDU-2803?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
HeLifu updated KUDU-2803:
-------------------------
Description:
We know that kudu does not allow the type of a column to be altered right now.
But indeed, a very common case is RDMS -> kudu, and there will be type
alterations inevitably, especially for small to large types.
Currently, here are two ways to get rid of this problem:
# for the new tables: predefine a large type for every column;
# for the existing tables: stop app(write) > add a new column with large type
> copy data -> drop old column -> rename new column to old column -> start app;
However, neither of them is elegant. So, i think it is necessary to support
non-primary key type alteration.
was:
We know that kudu does not allow the type of a column to be altered right now.
But indeed, a very common case is RDMS -> kudu, and there will be type
alterations inevitably.
Currently, here are two ways to get rid of this problem:
# for the new tables: predefine a big type for every column;
# for the existing tables: stop app(write) > add a new column with new type >
copy data -> drop old column -> rename new column to old column -> start app;
However, neither of them is elegant. So, i think it is necessary to support
non-primary key type alteration.
> support non-primary key type alteration
> ---------------------------------------
>
> Key: KUDU-2803
> URL: https://issues.apache.org/jira/browse/KUDU-2803
> Project: Kudu
> Issue Type: Improvement
> Reporter: HeLifu
> Priority: Major
>
> We know that kudu does not allow the type of a column to be altered right
> now. But indeed, a very common case is RDMS -> kudu, and there will be type
> alterations inevitably, especially for small to large types.
> Currently, here are two ways to get rid of this problem:
> # for the new tables: predefine a large type for every column;
> # for the existing tables: stop app(write) > add a new column with large
> type > copy data -> drop old column -> rename new column to old column ->
> start app;
> However, neither of them is elegant. So, i think it is necessary to support
> non-primary key type alteration.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
