[
https://issues.apache.org/jira/browse/HIVE-26555?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Teddy Choi updated HIVE-26555:
------------------------------
Description:
h1. Purpose
In failover/fail-back scenarios, a Hive database needs to be read-only, while
other one is writable to keep a single source of truth.
h1. User-Facing Changes
Yes. EnforceReadOnlyDatabaseHook class implements ExecuteWithHookContext
interface. hive.exec.pre.hooks needs to have the class name to initiate an
instance. The "readonly" database property can be configured to turn it on and
off.
h2. Allowed read operations
All read operations without any data/metadata change are allowed.
* EXPLAIN
* USE(or SWITCHDATABASE)
* REPLDUMP
* REPLSTATUS
* EXPORT
* KILL_QUERY
* DESC prefix
* SHOW prefix
* QUERY with SELECT or EXPLAIN. INSERT, DELETE, UPDATE are disallowed.
h2. Allowed write operations
Most of write operations that change data/metadata are disallowed. There are
few allowed exceptions. The first one is alter database to make a database
writable. The second one is replication load to load a dumped database.
* ALTER DATABASE db_name SET DBPROPERTIES without "readonly"="true".
* REPLLOAD
h1. Tests
* read_only_hook.q: USE, SHOW, DESC, DESCRIBE, EXPLAIN, SELECT
* read_only_delete.q
* read_only_insert.q
was:
h1. Purpose
In failover/fail-back scenarios, a Hive database needs to be read-only, while
other one is writable to keep a single source of truth.
h1. Design
Yes. EnforceReadOnlyDatabaseHook class implements ExecuteWithHookContext
interface. hive.exec.pre.hooks needs to have the class name to initiate an
instance. The "readonly" database property can be configured to turn it on and
off.
Allowed operations prefixes
* EXPLAIN
* USE(or SWITCHDATABASE)
* REPLDUMP
* REPLSTATUS
* EXPORT
* KILL_QUERY
* DESC
* SHOW
h1. Tests
* read_only_hook.q: USE, SHOW, DESC, DESCRIBE, EXPLAIN, SELECT
* read_only_delete.q
* read_only_insert.q
> Read-only mode for Hive database
> --------------------------------
>
> Key: HIVE-26555
> URL: https://issues.apache.org/jira/browse/HIVE-26555
> Project: Hive
> Issue Type: New Feature
> Reporter: Teddy Choi
> Assignee: Teddy Choi
> Priority: Minor
> Labels: pull-request-available
> Fix For: 4.0.0-alpha-2
>
> Time Spent: 40m
> Remaining Estimate: 0h
>
> h1. Purpose
> In failover/fail-back scenarios, a Hive database needs to be read-only, while
> other one is writable to keep a single source of truth.
> h1. User-Facing Changes
> Yes. EnforceReadOnlyDatabaseHook class implements ExecuteWithHookContext
> interface. hive.exec.pre.hooks needs to have the class name to initiate an
> instance. The "readonly" database property can be configured to turn it on
> and off.
> h2. Allowed read operations
> All read operations without any data/metadata change are allowed.
> * EXPLAIN
> * USE(or SWITCHDATABASE)
> * REPLDUMP
> * REPLSTATUS
> * EXPORT
> * KILL_QUERY
> * DESC prefix
> * SHOW prefix
> * QUERY with SELECT or EXPLAIN. INSERT, DELETE, UPDATE are disallowed.
> h2. Allowed write operations
> Most of write operations that change data/metadata are disallowed. There are
> few allowed exceptions. The first one is alter database to make a database
> writable. The second one is replication load to load a dumped database.
> * ALTER DATABASE db_name SET DBPROPERTIES without "readonly"="true".
> * REPLLOAD
> h1. Tests
> * read_only_hook.q: USE, SHOW, DESC, DESCRIBE, EXPLAIN, SELECT
> * read_only_delete.q
> * read_only_insert.q
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
