P3
ADDITIONAL SYSTEM INFORMATION :
We are using Openshift with the openjdk-21-ubi9:latest as base image.
A DESCRIPTION OF THE PROBLEM :
Since the update of the package from OpenJDK-21.0.3 to OpenJDK-21.0.4, we are experiencing that our pods increase at a steady pace, depending on the app/service is faster. The update of the JDK was the only change.
Sometimes the GC is not triggered and crashes the pods, other times we see the memory behavior go up until GC is triggered, and go down and start again to go up. Before memory was almost constant.
We could reproduce it in different apps.
REGRESSION : Last worked in version 21
STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Our apps are spring-boot in an Openshift cluster with openjdk-21-ubi9:latest as the base image. But probably you can check it locally also, with a simple app might take some time to see the increase, with an app with Kafka Streams you could detect the diference
EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
The expected result will be to have a quite steady heap memory.
ACTUAL -
As mentioned memory goes up and down, the garbage collector does not seem to work as before. If you let the app without a load in keeps increasing Heap memory while before was almost constant.
CUSTOMER SUBMITTED WORKAROUND :
We found 2 workarounds:
- use OpenJDK-21.0.3
- use the ZGC(-XX:+UseZGC -XX:+ZGenerational), that seems to still work fine
FREQUENCY : always
We are using Openshift with the openjdk-21-ubi9:latest as base image.
A DESCRIPTION OF THE PROBLEM :
Since the update of the package from OpenJDK-21.0.3 to OpenJDK-21.0.4, we are experiencing that our pods increase at a steady pace, depending on the app/service is faster. The update of the JDK was the only change.
Sometimes the GC is not triggered and crashes the pods, other times we see the memory behavior go up until GC is triggered, and go down and start again to go up. Before memory was almost constant.
We could reproduce it in different apps.
REGRESSION : Last worked in version 21
STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Our apps are spring-boot in an Openshift cluster with openjdk-21-ubi9:latest as the base image. But probably you can check it locally also, with a simple app might take some time to see the increase, with an app with Kafka Streams you could detect the diference
EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
The expected result will be to have a quite steady heap memory.
ACTUAL -
As mentioned memory goes up and down, the garbage collector does not seem to work as before. If you let the app without a load in keeps increasing Heap memory while before was almost constant.
CUSTOMER SUBMITTED WORKAROUND :
We found 2 workarounds:
- use OpenJDK-21.0.3
- use the ZGC(-XX:+UseZGC -XX:+ZGenerational), that seems to still work fine
FREQUENCY : always