[jira] [Commented] (AMBARI-14714) [Umbrella] Multi Everything Architecture

["Rachid BERKANE (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AMBARI-14714?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17107049#comment-17107049 ] 

Rachid BERKANE commented on AMBARI-14714:
-----------------------------------------

[~jluniya] [~sposetti]

some design suggestions:

one database per cluster:
 * may simplify cluster move (or configure DB HA for specific cluster)
 * simplify stacks upgrade dependencies mgmt ... ( different DBs version,  ...)
 * administration delegation: a lightweight ambari may allow to monitor or restart services, + ambari views
 * attach/detach a cluster from one ambari server to another or install/upgrade ambari independently then attache cluster to new ambari server

common/shared database: ambari clusters mgmt
 * link clusters : storage/compute/data-flow clusters : multi versionned compute clusters may use same storage server
 * manage hosts and bind ambari agents to specific clusters
 * migrate a cluster database to elsewere
 * possibility to manage viewfs/federation config for some clusters (add a hdfs DIR into a second clustes hdfs viewsfs)
 * replications config
 * inter-cluster resources federation
 * mpack repositories management

 

> [Umbrella] Multi Everything Architecture
> ----------------------------------------
>
>                 Key: AMBARI-14714
>                 URL: https://issues.apache.org/jira/browse/AMBARI-14714
>             Project: Ambari
>          Issue Type: Epic
>          Components: ambari-agent, ambari-server, ambari-upgrade, ambari-web, stacks
>    Affects Versions: 2.5.0
>            Reporter: Jeff Sposetti
>            Assignee: Jayush Luniya
>            Priority: Critical
>              Labels: pull-request-available
>             Fix For: 3.0.0
>
>         Attachments: ApacheAmbariMulti-EverythingArchitecture.pdf
>
>          Time Spent: 9h
>  Remaining Estimate: 0h
>
> *Multi Stack Services* (AMBARI-19621)
> _Scenario: Deploy HDP & HDF services in same cluster_
> - Deploy HDFS from HDP and Kafka, Storm, NiFI from HDF in the same cluster.
> _Scenario: Deploy HDP services from different stack versions in same cluster_
> - Deploy ZK, HDFS, YARN from HDP-2.5, but deploy latest SPARK from HDP-2.6.
> *Multiple Service Instances* (AMBARI-20463)
> _Scenario: Multi service instances on same version_
> - Cluster includes instance of ZooKeeper vX which is being used by HDFS, YARN.
> - User wants to add instance of ZooKeeper vX which is being used by STORM and KAFKA
> _Scenario: Multi service instances on different versions_
> - Cluster includes instance of SPARK vX.
> - User wants to add additional instance of SPARK vY.
> *Multi Host Component Instances* (AMBARI-20465)
> _Scenario: Multi component instances from a service instance on the same host_
> - Single host with 128GB RAM, 16 (actual) Cores, 12*4TB disks.
> - Single instance of KAFKA broker is unable to utilize all the resources on the host. 
> - User wants to scale up performance by deploying multiple instances of the KAFKA brokers/host.
> *Multi Cluster* (AMBARI-20466)
> _Scenario: Manage multiple Hadoop clusters under single Ambari Server_
> - Customer has multiple small Hadoop clusters and would like to manage and monitor them with a single Ambari Server instance.
> *Multi Yarn Hosted Services* (AMBARI-17353)
> _Scenario: HBase on YARN_
> - Deploy second instance of HBase as a long running YARN service.
> - Manage YARN hosted service similar to traditional hosted services. 
> - First class support for Yarn hosted services.
> _Scenario: Credit Fraud Detection YARN Assembly_
> - YARN Assembly can have its own ZK, KAFKA etc.
> - Manage YARN Assemblies as first-class citizen.
> *Multi Everything Aware APIs* (AMBARI-19620)
> - V2 Ambari Rest API to support Multi Everything Architecture



--
This message was sent by Atlassian Jira
(v8.3.4#803005)