Support Questions
Find answers, ask questions, and share your expertise
Announcements
Check out our newest addition to the community, the Cloudera Innovation Accelerator group hub.

Impala Client Connections

Explorer

Hello All,

 

We have been receiving alerts of IMPALAD_FRONTEND_CONNECTIONS becoming bad "There are 0 (Beeswax pool) 55 (Hive Server 2 pool) active client connections, each pool has a configured maximum of 64"

 

Impala Daemon Concurrent Client Connections Monitoring Percentage Thresholds
Warning = 80
Critical = 95

 

Impala Daemon Max Client Connections = 64

 

CM - 5.16.2
CDH - 5.16.2

 

Will increasing Impala Daemon Max Client Connections reduce / stop these alerts? Any help is most welcome.

3 REPLIES 3

Guru
@Amn_468

Having lots of connections to impala daemon might overload it and cause issues, this is the reason for the alert. Now the question is will your impala daemon handle those connections. If yes, you can increase the max client connections, but if no, we might need to find out why there are many connections that makes you about to go to the limit. Are there any connections that are supposed to be closed but still open?

You should check the resources being used on the host to determine if impala is really under pressure or not and decide your action.

Cheers
Eric

Explorer

Hi Eric,

 

Thanks for your reply, post your reply, I have more questions than answers 🙂

1)  Is there a way to find if impala deamon can handle additional connections (from default 64 to 128 and more)

2) Can we find the number of connections that are made in real time? (Apart from Active Frontend API Connections Chart)

3) How are connection closed (i.e. are they closed once an impala query is completed or is there some other process)

4) Any best practice recommendation for Swap Memory setting in Impala.

 

Regards

Amn

New Contributor

Seems like your cluster is configured with default impala admission control control. Please create and setup new impala admission control control with the correct parameter based on the how much queries need to executed and resources you have