[
https://issues.apache.org/jira/browse/RANGER-3771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17545358#comment-17545358
]
Binhua Hu commented on RANGER-3771:
-----------------------------------
The difference from deploying Ranger in non-HA mode is:
* Deploy RANGER_ADMIN component to multiple host
* Setup a load balancer and configure it to front all RANGER_ADMIN instances
(The URL of a Ranger Admin instance is [http://host:port|http://hostport/]
(default port 6080) )
* admin-properties
** policymgr_external_url - override the value of this configuration property
with the URL of the load balancer. Each component interacting with Ranger is
using the value of this property to connect to Ranger thus these will connect
via the balancer.
For example, in the k8s environment, the k8s entry agent can be used as a load
balancer.
> The ranger rest client can easily overload a single server in HA without
> load-balancer
> --------------------------------------------------------------------------------------
>
> Key: RANGER-3771
> URL: https://issues.apache.org/jira/browse/RANGER-3771
> Project: Ranger
> Issue Type: Improvement
> Components: Ranger
> Affects Versions: 2.1.0
> Reporter: Binhua Hu
> Priority: Major
>
> The rest client used in the ranger source code(RangerRESTClient) only
> supports high availability and has no load balancing strategy, which is easy
> to lead to hot issues.
> RANGER-2555 also mentioned. According to my understanding, multi node
> rangeradmin should be able to provide services at the same time. So why not
> add a load balancing policy？
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
