Closed
Description
Before reporting an issue
- I have read and understood the above terms for submitting issues, and I understand that my issue may be closed without action if I do not follow them.
Area
dist/quarkus
Describe the bug
Main tracking issue: defenseunicorns/uds-core#1212
When using Fedora F41 (kernel 6.12 line), Keycloak Pod get OOM Killed. The root cause can clearly be seen when manually calling the following command from the Keycloak Pod:
Kernel 6.11 (Good):
sh-5.1$ java -XshowSettings:vm -version
Max. Heap Size (Estimated): 247.50M <-- !!!
...
Provider: cgroupv2
Effective CPU Count: 1 <-- !!!
Kernel 6.12 (Bad):
sh-5.1$ java -XshowSettings:vm -version
...
Max. Heap Size (Estimated): 7.67G <-- !!! This is the Node value!
...
Provider: cgroupv2
Effective CPU Count: 16 <-- !!! Again, taken from the Node!
The issue has already been reported in JDK-8346874 that follows to the actual fix - JDK-8347129. At the time of writing, this has been fixed only in the latest JVM versions with no information about the backports
Version
26.0.8
Regression
- The issue is a regression
Expected behavior
Memory should be properly interpreted
Actual behavior
Memory is being pulled from the Node running the container
How to Reproduce?
Deploy Keycloak on Fedora F41 (any Kubernetes distribution that doesn't use Virtual Machines, for example k3d) with Kernel 6.12 onwards
Anything else?
No response