accumulo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Eric Newton <eric.new...@gmail.com>
Subject Re: client.conf warning in 1.7.0
Date Wed, 01 Jul 2015 15:02:03 GMT
The ticket for documentation exists:

https://issues.apache.org/jira/browse/ACCUMULO-3851

I'm going to unassign it from myself. I think it's important to document,
but I had nothing to do with it's implementation, so I'll let someone more
knowledgeable tackle it.



On Tue, Jun 30, 2015 at 11:01 PM, Christopher <ctubbsii@apache.org> wrote:

> That's correct. And, that's precisely the kind of things it'd be nice to
> add. I would say that it's on the agenda.
>
> On Tue, Jun 30, 2015, 21:08 Max Thomas <max.thomas@jhu.edu> wrote:
>
>> Christopher/Josh - thanks for the quick responses.
>>
>> I certainly understand the concerns about passing around
>> accumulo-site.xml. While I haven't had any issues with this, based on
>> poking around ClientConfiguration.java it seems that it was necessary
>> for various SSL/Kerberos features.
>>
>> Forgive what might be a very naive question, but: currently, as a
>> consumer of an Accumulo from Java, I've got to provide (minimally) a
>> username, password, instance name, and zookeeper server string. Based on
>> a brief scan of ClientConfiguration, though, it does not seem that the
>> username/password will be read from client.conf - is that correct? Will
>> support for that be added in the future?
>>
>> On 6/30/15 8:07 PM, Josh Elser wrote:
>> > I'll open up a documentation ticket. I thought we had something in there
>> > but maybe not.
>> >
>> > For a normal instance, this file would likely just contain
>> >
>> > instance.zookeeper.host=zkserver1,zkserver2,etc..
>> >
>> > Like Christopher says, it's not required yet, but it would be a good
>> > habit to start using it as it's the direction we're heading.
>> >
>> > Christopher wrote:
>> >> In short, yes, it will likely be required for clients in the future
>> >> (if you want something other than the defaults).
>> >>
>> >> The longer answer is that we just wanted to make sure the user
>> >> understood where their configuration is coming from, in case things
>> >> aren't working as they expect.
>> >>
>> >> We've been moving towards separating the server-side configuration
>> >> (mostly, accumulo-site.xml) from client utilities, due to the fact
>> >> that a client cannot be expected to have access to the
>> >> accumulo-site.xml (it contains a sensitive instance.secret and other
>> >> stuff a client doesn't need), and it can be confusing to have a
>> >> client-specific "accumulo-site.xml", which is not the server's
>> >> "accumulo-site.xml".
>> >>
>> >> If you don't have this file, it will fall back to reasonable defaults
>> >> (reading "accumulo-site.xml" from the classpath, if one is available).
>> >>
>> >> Some of this behavior is a bit screwy, but we've been careful to
>> >> preserve backwards compatibility as much as possible. In the future,
>> >> it is my intention (and perhaps others share this goal with me) to
>> >> clean up these configs quite a bit. So, if you have any suggestions on
>> >> how you think things should work, feedback is always welcome.
>> >>
>> >> --
>> >> Christopher L Tubbs II
>> >> http://gravatar.com/ctubbsii
>> >>
>> >>
>> >> On Tue, Jun 30, 2015 at 7:37 PM, Max Thomas<max.thomas@jhu.edu>
>> wrote:
>> >>> When running code against a 1.7.0 cluster, the following warning keeps
>> >>> appearing:
>> >>>
>> >>> 2015-06-30 19:32:58,352 WARN  o.a.a.c.c.ClientConfiguration [main]
>> >>> Found no
>> >>> client.conf in default paths. Using default client configuration
>> values.
>> >>>
>> >>> The only mention of this file that I can find, however, is in the
>> manual
>> >>> buried in a section on Kerberos.
>> >>>
>> >>> Is this a warning because this will be required for clients in the
>> >>> future?
>>
>

Mime
View raw message