I was planning on working ACCUMULO-135 for 1.5.  I suppose in the
process of doing this that I could also generalize the authentication
mechanism.

On Thu, Jan 5, 2012 at 1:19 PM, John W Vines <john.w.vi...@ugov.gov> wrote:
> We've been mumbling about making the authentication system more pluggable. 
> Right now one of the standing issues is that we have the credentials need to 
> perform actions a bit tightly woven with the ZKAuthenticator. These need to 
> be segregated better before more progress can be made allow more 
> authenticators.
>
> John
>
> ----- Original Message -----
> | From: "Eric Newton" <eric.new...@gmail.com>
> | To: accumulo-dev@incubator.apache.org
> | Sent: Thursday, January 5, 2012 9:27:18 AM
> | Subject: Re: zookeeper ACL issues
> | The contents of the user node is the 8-byte salt and salted SHA-256 of
> | the
> | user's password. I don't believe there's any encoding: it's just raw
> | bytes.
> |
> | We have not done anything to extend the authentication system, yet.
> | Hopefully we'll be able to leverage the tools now being released for
> | the
> | rest of the Hadoop infrastructure.
> |
> | -Eric
> |
> |
> | On Wed, Jan 4, 2012 at 9:14 PM, Jim Klucar <klu...@psualum.com> wrote:
> |
> | > That worked for the CLI. What is the data in there? base64 encoding
> | > of the
> | > password?
> | >
> | > Have you heard of anyone trying to implement an OAuth style
> | > authentication
> | > for the Authenticator? I was thinking I would have to put in a whole
> | > layer
> | > that talked to ZK directly for authentication, but perhaps not.
> | >
> | > thanks for help.
> | >
> | > On Wed, Jan 4, 2012 at 3:48 PM, Eric Newton <eric.new...@gmail.com>
> | > wrote:
> | >
> | > > In fact, that's why there is an abstraction for the Authenticator,
> | > > hopefully you can replace it with one of your own.
> | > >
> | > > However, here's the trick for reading the data
> | > >
> | > > zkCli > addauth digest accumulo:DEFAULT
> | > >
> | > > If you have changed the value of "instance.secret" in
> | > > accumulo-site.xml,
> | > > you need to use that value in order to read what is in the node.
> | > >
> | > > The data there is binary, so you'll need to write some java code
> | > > to
> | > decode
> | > > it... if your process can read accumulo-site.xml, you can use
> | > > ZooReaderWriter to pull the data out: it will always set the
> | > > additional
> | > > auth flags to be able to read those nodes.
> | > >
> | > > Let me know if this doesn't make sense, and I can provide more
> | > > details.
> | > >
> | > > -Eric
> | > >
> | > > On Wed, Jan 4, 2012 at 3:09 PM, Jim Klucar <klu...@psualum.com>
> | > > wrote:
> | > >
> | > > > Hey,
> | > > >
> | > > > I've been poking at security features of Accumulo, specifically
> | > > > the
> | > > > user/password stuff in Zookeeper. I was wondering if it would be
> | > possible
> | > > > to abstract away the zookeeper ACL authentication, so we could
> | > > > deploy
> | > > with
> | > > > our own instead of using the "digest" ACL stuff in zookeeper. I
> | > > > looked
> | > at
> | > > > doing this but got lost a bit in the innards of how Accumulo
> | > > authenticates
> | > > > the znodes. Can anyone provide me some guidance? Specifically,
> | > > > it
> | > would
> | > > be
> | > > > useful to know what to do to be able to do an ls of the
> | > > > /accumulo/<instance>/users/<user> zookeeper path from zkCli.sh
> | > > >
> | > > > Thanks,
> | > > > Jim
> | > > >
> | > >
> | >

Reply via email to