incubator-hcatalog-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ashutosh Chauhan <hashut...@apache.org>
Subject Re: Permissions and Ownership for Table Data
Date Mon, 01 Aug 2011 17:57:21 GMT
> So there's no meta-data that stores the permissions/group for tables.
> All we need to do then is to make sure that the warehouse directory in
> HDFS for the table has the right permissions and group ownership and
> that we continue to load data via HCatOutputFormat.
>
> Right?

Right. HCatalog doesn't store any metadata w.r.t permissions/groups.
What you said is precisely how we envision HCatalog to be used.

Ashutosh
On Sun, Jul 31, 2011 at 21:16, Ranjit Mathew <ranjit@yahoo-inc.com> wrote:
> On 07/29/2011 05:48 AM, Ashutosh Chauhan wrote:
>>
>> Sorry for bit late on this. When you add new partitions permissions
>> and groups do get inherited from table, if you are adding your data
>> through HCatLoader (via pig) or through HCatOutputFormat (via your
>> map-reduce program). If you are adding data through other mechanism
>> and then doing "add partition" on CLI, in that case those won't get
>> inherited.
>
> Thanks for clarifying this Ashutosh. We *are* loading our data via
> HCatOutputFormat.
>
> So there's no meta-data that stores the permissions/group for tables.
> All we need to do then is to make sure that the warehouse directory in
> HDFS for the table has the right permissions and group ownership and
> that we continue to load data via HCatOutputFormat.
>
> Right?
>
>
>> AFAIK, you can't set groups/perms through Hive CLI.
>
> Sort of - you can use the "dfs" command to do it (that's how we're
> doing it right now).
>
> Thanks,
> Ranjit
>
>
>> On Tue, Jul 26, 2011 at 04:33, Ranjit Mathew<ranjit@yahoo-inc.com>  wrote:
>>>
>>> Hi,
>>>
>>>  We are trying to make sure that for a table "snafu", data
>>> is owned by group "foo" for a user "bar" and has permissions "750".
>>>
>>> The HCatalog CLI program seems to provide "-p" and "-g" options
>>> just for this purpose. However, this only seems to set the
>>> permissions and group-ownership for the location in HDFS containing
>>> the table-data (say "/wombat/snafu") at the time of table-creation.
>>>
>>> What we are looking for is that as the data is populated into the
>>> table, it continues to be owned by "bar" belonging to "foo" and
>>> has the permissions "750". How can we ensure this?
>>>
>>> FWIW, we're using HCatOutputFormat to directly write data into
>>> this table. We're using HCatalog 0.1 with Hadoop 0.20.204.
>>>
>>> If the table has already been created using Hive CLI (not HCatalog CLI),
>>> is there a way to set its group/permissions attributes? (We don't
>>> want to drop and recreate the table as it has a lot of data and continues
>>> to be populated.)
>>>
>>> Thanks in advance for your help,
>>> Ranjit
>>>
>
>

Mime
View raw message