After installing Wily Agent 2023.5.1.13(Build 990013) on the XD1 system, the customer observed high memory and CPU load on application nodes. After second restart, one node persists with having high memory and CPU load.
Also server process shutdowns are occurring with errors in the logs ,example below
"Service process shutting down with exit code [666]; memory allocation error [OutOfMemoryError]"
The heap dump shows the Agent is overloaded with the AgentNameChange event handling:
Most of the heap is occupied by the execution queue for "Agent Execution" thread ("com.wily.util.task.AsynchExecutionQueue"), which contains huge number of entries, most of them are NotifyNameChange and AgentNameChangedExecutableItem.
Introscope Agent 2023.5.1.13(Build 990013)
JVM version of managed system XD1 : SAP JVM 8.1.095
The root cause is a bug on agent log re-naming introduced when we switched our logging lib from log4j to logback.
The workaround is set the below in IntroscopeAgent.profile file
introscope.agent.disableLogFileAutoNaming=true.
This disable log file name auto renaming feature will avoid an endless loop trying to resolve log file name, which keeps consuming memory.
The issue is fixed in which we've fixed in later releases (24.1.1+)