Is there any HA feature available for hive in MAPR?
I would also take a look at this JIRA. [HIVE-8376] Umbrella Jira for HiveServer2 dynamic service discovery - ASF JIRA
I think they are working on a built in version of HS2 proxy type config. Right now you do this with HAProxy which has detailed configurations available to control how the job is routed. The metastore HA is done however the database being used normally would do HA. MySQL for example has many different methods.
You could start multiple HS2 and Hive metastore services on different nodes or even the same nodes, but your application layer or loadbalancer needs to be configured properly.
This is more like a Active-Active HA.
Thanks Hao, if iam not wrong so Active-Active HA is typically a Load Balancing Interface rather than High Availbility. Do you have any Documentation to deploy active-active HA. Also any timeline for this feature to be included as part of enterprise MAPR edition?
You can just install and start HS2 or HM on multiple nodes or on the same nodes listening on different ports.
Is this what you want? Or could you give more details on the feature request?
So if I install HMS(New) and HS2(New) on Separate node
1)how the existing/new Client connections are Routed to HMS(New/Old) and HS2(New/Old) instances ?
since we are not using any loadbalancer , just installing and updating configs
My Feature Request would be like More Like an active-standby for HMS just like CLDB HA ?
There is no way right now to auto-route Hive queries to different nodes.
Application layer or LoadBalancer is needed for this.
Regarding the feature request, let me Neeraja Rentachintala for any inputs.
Thanks Hao, what load balancer do you suggest nginx or haproxy or anyother .
Appreciate your effort for considering my feature request.
I do not have preference. But I heard HAproxy a lot to do this.
Also we achive true active-standby HA only if the underlying Metastore Database mysql or postgres is configured with HA.
That is true and I believe that is more important.
For Mysql, you can consider MySQL Cluster;
For Postgres, Greenplum is another choice.
Thanks Hao MYSQL Cluster , Green Plum are they open source or bound by enterprise Licence?
Hiveserver2 dynamic service discovery not working in MapR Environment.
Retrieving data ...