Support Questions

Find answers, ask questions, and share your expertise

JDBC Connection String with multiple worker node Impala

avatar
New Contributor

Hi All,

 

We are working to connect our java application to the multiple nodes of impala worker using the JDBC connection string, right now we connected to use only single node "worker-01" if this is down the whole connection will down.

 

Is there any way to connect to multiple worker nodes using one JDBC connection string?, this connection string having a Kerberos authentication along with SSL for each node. that mean if i need to connect to different worker node i need to change the certificate along with the KrbFQDN. 

3 REPLIES 3

avatar
New Contributor

currently we are using the cloudera JDBC driver to connect to the cloudera platform. Also if the driver support to have the multiple connection, then please provide the valid format that will be supported by the driver. 

avatar
Expert Contributor

Hello @bb9900m 

 

The use case in your scenario is make use of Load balancer. You can make user of external Load Balancer and balance the impala coordinators behind the Load balancer.

 

This way, you will have the Load balancer Ip exposed to the client(Virtual IP) which further will balance the load based upon the load balancing mechanism set.

 

Attaching the link and sample configuration for your reference.

 

https://docs.cloudera.com/runtime/7.2.6/impala-manage/topics/impala-load-balancer-configure.html

 

When it comes to SSL, make sure you have the LB hostname added as SAN names on all the impala certificates.

 

Let me know if you have issues in any of the points above.

avatar
New Contributor

Thanks for your reply @tusharkathpal,

 

We already addressed this case with our customer to have the load balancer and vIP. but the customers seeing this is a risk points to expose the worker node IPs to the external departments within the same Organization and needs to be handled internally within the same application ~ department.

 

Is there any alternative scenarios that can be achieve the same approach without exposing the worker node IPs to external system like F5?