accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Keith Turner (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-2076) Make it easier to connect to acccumulo
Date Thu, 02 Jan 2014 14:45:52 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-2076?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13860232#comment-13860232
] 

Keith Turner commented on ACCUMULO-2076:
----------------------------------------

bq.  The ZooKeeperInstance(Configuration) constructor is still too much work, I think. The
user still has to feed an instance name/id into the connection. 

If the user creates a file containing the following props and passes that file to the zookeeper
instance constructor, then they can connect w/o specifying the instance name programmaticly.

{noformat}
client.instance.name=<instance name>
instance.zookeeper.host=<zookeepers>
{noformat}

Given that a generic Apache commons config object is passed into the constructor, we could
probably read xml formatted config. 

In discussion on ACCUMULO-1771, [~ctubbsii] suggested specifying the instance name in the
server config.   Given this suggestion it would probably be wise to change client.instance.name
to instance.name, so that the same prop could be used in client and server config.  If this
were done, then we could probably make the client and server easily use the same config file
if someone wanted to.   I suppose the only missing piece is finding the config file on the
classpath.  I would recommend modifying ClientConfiguration to have a method to do this rather
than creating a new class.    ClientConfiguration already has some methods to aid users in
finding config files.    Adding other classes that help users find config files is confusing
when consider the entire API.

> Make it easier to connect to acccumulo
> --------------------------------------
>
>                 Key: ACCUMULO-2076
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-2076
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: client
>            Reporter: Mike Drob
>            Priority: Minor
>             Fix For: 1.7.0
>
>
> Given a program that has {{accumulo-site.xml}} on its classpath, we should be able to
abstract away from the user the steps for connecting to accumulo. I'm thinking that we need
to expose something like HdfsZooInstance through the client API, but not exactly that, since
HZI looks like it carries some extra functionality.
> # Determine the backing file system
> # Get Instance ID from the FS. 
> # Get ZooServers from accumulo-site.xml
> # Get Instance Name from ZK.
> # Create an Instance and hand it to the user.
> If this already exists in master, then maybe we need to do a better job of documenting.
Or a better job of educating me.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Mime
View raw message