Created on 07-12-2017 04:49 AM - edited 07-12-2017 05:33 AM
Hi,
I have created F5 VIP for the impala nodes in my cluster that helps me to distribute the query load on the nodes.
When i'm using the VIP in the impala-shell all is working fine and all the queries distibute on the different impala nodes.
When i tried to configure it in Hue by adding the following to the Hue snppest valve:
[impala]
server_host=impala-a.init.liveperson.net
server_port=21050
server_conn_timeout=3600
I got an error that the node is down or the metasore is down.
I tried also to add the vip in the impala loader balancer configuration and got the same error.
Maybe i need to open the port from the Hue server to vip?
Any idea how to configure this?
Created 07-12-2017 05:14 PM
Are session sets as Sticky?
https://www.cloudera.com/documentation/enterprise/5-10-x/topics/impala_proxy.html
Consider enabling "sticky sessions". Cloudera recommends enabling this setting so that stateless client applications such as impalad and Hue are not disconnected from long-running queries. Evaluate whether this setting is appropriate for your combination of workload and client applications.
Note: in next CDH release, Hue should handle the Impala load balancing out of the box.
Created 07-12-2017 05:19 PM
Created 07-12-2017 08:33 PM
@Romainr The documntation you provided is for the HA proxy server while as i mentioned i'm using the F5 VIP and already read the documntation of using F5 for load balancing:
http://www.cloudera.com/documentation/other/reference-architecture/PDF/Impala-HA-with-F5-BIG-IP.pdf
Created 07-20-2017 10:02 PM
Managed to solve the issue by just adding the balancing configuration.
Using the VIP created a problem for users how drop and create tables where the impala-meta data was not syned cross all the nodes, i'm followed this link to get this work and currently testing it.
Wondering when this will be solved without using workarounds.
Created 07-21-2017 05:50 PM