Memory eaten up by java processes

[VinodNN]

Hello Erman,

From last few days we are having this recurring problem where memory on the apps node is hogged by java processes. After taking restart everything becomes normal but again in two-three days the RAM is used up and SWAP usage also starts climbing. Average SWAP usage for this instance has been around 35 percent, but now it is going above 50-60 and still higher, leading to hang situation.

How can I find out why the java processes on apps tier are behaving this way?

Regards,
Vinod

[ErmanArslansOracleBlog]

1)EBS apps tier has several Java processes. Don't chase "Java" as a whole — find the specific managed server.
Identify it. Use ps command.. something like ->   ps -eo pid,user,rss,vsz,etime,cmd --sort=-rss | grep -iE "java|weblogic" | head -20

2)confirm the box isn't simply over-committed. Look at your context file ($CONTEXT_FILE) for:

grep -E "s_oacore_nprocs|s_oacore_jvm_start_options|s_oacore_jvm_main_options|s_forms_jvm_start_options|s_oafm_jvm_start_options" $CONTEXT_FILE

3)
Compute: nprocs × Xmx per service, sum across services. If total is close to physical RAM + you also have CM, Listener, OS cache, an OS-level memory pressure is expected before any leak.

  free -h
  sar -r 1 1     # current
  sar -r -s 09:00:00 -e 18:00:00   # business-hours history
  sar -W         # swap activity over time (pswpin/pswpout)
  vmstat 5 10    # watch si/so columns

4)
In parallel open a SR - >with OS logs + the heap dump + GC log; that's enough for Oracle Support to triage immediately