top oracle server process not in v$process.spid
- From: Dba DBA <oracledbaquestions@xxxxxxxxx>
- To: ORACLE-L <oracle-l@xxxxxxxxxxxxx>
- Date: Tue, 9 Sep 2014 13:23:41 -0400
11.2.0.4 , 2 node rac
linux 6
AWR/ASH not turned on. Statspack not installed. Some people dont pay for
this.... tracking this down is tricky.
ran top, 2 oracle processes at 99% cpu. However, 0% memory. Do not show up
in gv$process.spid or setospid (for hanganalyze/system state dump)
I believe I have seen high cpu processes before that dont show up either,
but I usually just go to ASH or awr and find it easily. I don't have that.
Plus they dont appear to use connectin pooling, so
gv$session_event,gv$session_wait are not useful... so history.
Step 1:
Process So far: Top , shows two oracle session at 99% CPU
ps -ef | grep <process>
oracle<DBNAME>1 (DESCRIPTION=(LOCAL=YES)(ADDRESS=(PROTOCOL=beq)))
step 2: select count(*) from gv$process where spdi = <mypid>
no rows returned.
step 3: system statedump
SQL > oradebug setospid <HIGH_CPU_PID>
ORA-00075: process "Unix process pid: <HIGH_CPU_PID>, image: <none>" not
found in this instance
step 3: general system statedump 'oradebug setospid myospid'
not much here. S ee this. Read through references to library cache, but
when I check gv$session_event does not seem high. However, lots
HM: Master is instance 1. My instance is 1 - I am the master.
Hang detection is enabled. Hang resolution scope is PROCESS.
HM: CPU load is NOT abnormally high.
HM: IO load is NOT abnormally high (1).
see references to librarcyh cache latch. However, sessions dont use
connectin pooling and vanish before I can check sessions.
step 4: doesnt appear to use conneciton pooling. Number of connetions goes
between 60 - 200. So gv$session_Event, gv$session_wait are not real
useful...
step 5: poll gv$session over and over and over again.
occassionally see vktm idle process, pmon timer, rdbms ipc message as top
waits. but they flicker off.
Other related posts:
