Describe the bug
We recently upgraded our crt library from 0.24.0 to 0.31.1 and noticed massive cpu spikes. It seems that the library tries to use all the cpu the machine has, which is beyond what the kube container is allocated. I've tried toggling maxConcurrency and targetThroughputInGbps but it has no affects. I've decreased the "apparent" cpu count in java by using -XX:ActiveProcessorCount=10. The very odd part is in datadog the cpu spikes seem to be from lock functions. Attached are two files during a load test. You can see the change in cpu time for the underlying library


Expected Behavior
Should not utilize all cpu available
Current Behavior
Utilizes all cpu available
Reproduction Steps
We see this in load tests, i am not sure how to reproduce it reliably
Possible Solution
No response
Additional Information/Context
No response
aws-crt-java version used
0.31.1
Java version used
17
Operating System and version
Ubuntu 20.04.6 x86_64
Describe the bug
We recently upgraded our crt library from 0.24.0 to 0.31.1 and noticed massive cpu spikes. It seems that the library tries to use all the cpu the machine has, which is beyond what the kube container is allocated. I've tried toggling
maxConcurrencyandtargetThroughputInGbpsbut it has no affects. I've decreased the "apparent" cpu count in java by using -XX:ActiveProcessorCount=10. The very odd part is in datadog the cpu spikes seem to be from lock functions. Attached are two files during a load test. You can see the change in cpu time for the underlying libraryExpected Behavior
Should not utilize all cpu available
Current Behavior
Utilizes all cpu available
Reproduction Steps
We see this in load tests, i am not sure how to reproduce it reliably
Possible Solution
No response
Additional Information/Context
No response
aws-crt-java version used
0.31.1
Java version used
17
Operating System and version
Ubuntu 20.04.6 x86_64