11gR2 Grid Infrastructure oraagent.bin or orarootagent.bin process uses high CPU, memory and threads. If the cluster node is up and running for very long time, it could exhaust memory and swap space which could result in node eviction or other stability issue. With OS "top" or other memory monitoring tool, we can see huge virtual memory is allocated and increasing over
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
7242 oracle 18 0 308g 953m 14m S 83.2 0.7 166:28.72 oraagent.bin
In above example, 308GB of virtual memory is allocated to the agent process while it's consuming 83.2% of CPU
It's recommended to apply latest patch set or PSU. But if patches are unavailable or can not be applied, the workaround is to kill agent process during off peak hour. There should be no impact as agent will be respawned automatically.
To kill agent process with pid 7242: kill -9 7242