[jira] [Reopened] (KYLIN-3865) Centralize the zookeeper related info

["Shaofeng SHI (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/KYLIN-3865?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Shaofeng SHI reopened KYLIN-3865:
---------------------------------

Looking at this commit [https://github.com/apache/kylin/commit/a6aa9c812aa7cf93d97fcd8eb554a8c289d1f20d,] I think a feature downgrade:

 

The origin:
{code:java}
JOB_ENGINE_LEADER_PATH = "/kylin/%s/job_engine/leader";
KYLIN_SERVICE_PATH = "/kylin/%s/service";
...

String jobEnginePath = getJobEnginePath(slickMetadataPrefix(kylinConfig.getMetadataUrlPrefix()));
{code}
now:
{code:java}
static final String JOB_ENGINE_LEADER_PATH = "/job_engine/leader";
static final String KYLIN_SERVICE_PATH = "/service";
...
String jobEnginePath = JOB_ENGINE_LEADER_PATH;{code}
This means, previously, Kylin support multiple Kylin clusters in one Hadoop/zk/hbase cluster (with different metadata url); After this change, this is not allowed, as the lock name is the same.

 

[~yaho] could you please check and give the comments? thx!

> Centralize the zookeeper related info
> -------------------------------------
>
>                 Key: KYLIN-3865
>                 URL: https://issues.apache.org/jira/browse/KYLIN-3865
>             Project: Kylin
>          Issue Type: Improvement
>            Reporter: Zhong Yanghong
>            Assignee: Zhong Yanghong
>            Priority: Major
>             Fix For: v3.0.0
>
>
> Currently zookeeper related infos, like zk path, curator client, are scattered many places, like 
> ZookeeperDistributedLock, DistributedScheduler, CuratorScheduler. It's better to centralize them. There'll be several benefits:
> * Avoid creating multiple curator clients
> * Easy to manage the zk path by leveraging zkChRoot



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)