Unable to create Two scalable, fault-tolerant Data servers running Historical and MiddleManager processes

Hello Guys,

We are new to druid & setup a Basic Single Node Druid Cluster. But if

We are trying to create Two scalable, fault-tolerant Data servers running Historical and Middle Manager processes

But we do not know how to tell cluster where is another Data Server

Please help

Thanks

Like IP Address of another Data Server Node

Hi ,

This link should help you to set up multiple data nodes/Historicals :
https://druid.apache.org/docs/latest/tutorials/cluster.html

You should run the druid in clustering mode as explained in the above link and can run any number of datanodes .

Thanks and Regards,

Vaibhav

The common run time property defines the zk hosts which is deployed across all machines. The ZK ensemble will then know what servers are in the cluster. The data server has a runtime property where it has druid.host attribute which is used to announced itself in the cluster to be reachable.

Rommel Garcia

Thanks Rommel,

I have created a cluster as per your suggestion.

But if we query druid_broker node than we do not receive data in synchronized way

i.e, we got multiple results.

Please tell me how to overcome this problem.

Many thanks.

Hi,
Can you give more details about “got multiple results” ?

For example, druid has groupBy which gives you exact results and this should be the same result for same query

and druid also has TopN which runs approximate algorithms and might give different results for same query.

Is that what you mean by “got multiple results” ?

If yes, then it is expected behavior within reasonable limits.

If NO, can you give us more details about getting multiple results from broker queries ?

Thanks,

–siva