VMSTAT ,cpu run queue and idle wait events,job delayed
Hi,
We have a 96 core server with 48 cpus and 2 cores per cpu so total number of cpu's is 96.When we run some jobs i noticed that the cpu run queue gets full i.e the r field of vmstat output shows that around 120 processes are waiting to get cpu even though only 96 cpus are there.The utilisation of cpu generally is
usr sys idle
30 50 20
But in database i see a lot of sqlnet message from client idle wait event which is causing the job to run slow.Is it because of CPU congestion or is there any other problem like network?
Also if a process goes idle waiting for user input(event:sqlnet message from client) after sometime say 1ms it gets the input from the user but it doesnot get the cpu what will be the wait event ?Will it be sqlnet message from client or some other event?
And when I check the glance output during heavy load all top processes are oracle processes and BLOCK ON column in glance shows a lot of SOCKT entries during high load what does it mean.