Our Community is getting an upgrade! To get everything ready for the relaunch, we’ll be placing the site in read-only mode starting September 21st.
We really appreciate your understanding while we get things set up behind the scenes. Catch up on all the exciting details about the move here.
Need help or have questions? Drop us a line at [email protected]
Created on 08-11-2017 01:44 AM - edited 09-16-2022 05:04 AM
As per the title really. I'm seeing high memory usage on a few data nodes ,which is down to impala, however at the time there are no active impala queieres running on the nodes.
If I restart the impalad the memory will clear and remaing free until a query runs. I can monitor the query using the memory, but when it finishes the memory is not released back to the system.
Created 08-11-2017 07:55 AM
It's normal for idle Impala daemons to hold onto 1-2GB of memory plus the JVM heap memory. Any more than that may indicate something wrong.
The memz debug page has diagnostics for this: http://impala-daemon:25000/memz?detailed=true . You can see there if a fragment of a query is holding onto memory.
There are two common ways this can happen:
* The query wasn't cancelled and closed. In this case it should show up on the /queries page of the coordinator
* The query was cancelled and closed, but a fragment continued running (this is a bug). In that case you should see a running fragment-execution thread in /threadz on the impala daemon web page.
Created 08-11-2017 06:15 AM
As far as I understand how Impala works, that is the expected behaviour.
It is indeed intended for speeding up later queries that use the same sets of data.
Created 08-11-2017 07:55 AM
It's normal for idle Impala daemons to hold onto 1-2GB of memory plus the JVM heap memory. Any more than that may indicate something wrong.
The memz debug page has diagnostics for this: http://impala-daemon:25000/memz?detailed=true . You can see there if a fragment of a query is holding onto memory.
There are two common ways this can happen:
* The query wasn't cancelled and closed. In this case it should show up on the /queries page of the coordinator
* The query was cancelled and closed, but a fragment continued running (this is a bug). In that case you should see a running fragment-execution thread in /threadz on the impala daemon web page.
Created 08-11-2017 07:57 AM
Thank you for the detailed answer