J. Ryan Earl 2011-06-17, 21:09
-Re: Apache-hadoop => CDH3 conversion: permissions
Konstantin Boudnik 2011-06-17, 21:23
I'm pretty sure you'll find more support on a cdh specific mailing lists.
Apparently, such a conversion won't be covered by Apache Hadoop documentation.
On Fri, Jun 17, 2011 at 04:09PM, J. Ryan Earl wrote:
> I'm trying to nail down a process for converting existing Apache-hadoop
> clusters with significant amounts of pre-existing data to CDH3. While
> I've found documentation for upgrading between CDH versions, I haven't
> seen one for standard Apache-hadoop => CDH3 with the "new" mapred and hdfs
> users and groups. I'm looking for the proper method to manually convert
> permissions on existing data from a single user&group setup
> (hadoop/hadoop) to the 2user & 3 group setup of hdfs/mapred users with
> hdfs/mapred/hadoop groups.
> What I'm thinking needs to happen is something like this:
> 1. Shutdown cluster.
> 2. Perform full configuration and HDFS data backup.
> 3. Delete existing hadoop user/group while leaving HDFS data/mapred
> folders untouched.
> 4. Install CDH3 packages (which sets up new users and groups).
> 5. Manually adjust permissions/groups/ownership of old data files to
> match new CDH3 security setup.
> 6. Flow into standard hadoop-upgrade process.
> Basically, I'm trying to nail down setup 5, and would appreciate any
> guidance on this. I'm guessing I just haven't found the correct document
> since this seems like it would be a common endeavor.
> Thanks in advance,