Regardless of OS, it seems that one PMON process was leftover from a
prior startup of the database instance. What did the UNIX "ps -ef"
command show for the start-time for each of the processes --- I'll bet
they were different. Even if they weren't different, if the starts
were prior to yesterday, then "ps" will show only the date not the
time, so they could have been from different startups within the same
day.
At any rate, it's not that unusual. It's good that you cleaned up the
shared memory still held by the prior instance startup -- have you
checked for the same thing on semaphores?
What causes it? Hmmmm..... often you have the situation where one
server process in an instance is waiting indefinitely for a system call
(i.e. I/O request) and thus it doesn't check to see if it should die
due to SHUTDOWN command. So, it persists. And since it has an open
file-handle on the shared memory, it holds that open too. Of course,
it could be due to a myriad of reasons, and it is good to chase it down
to understand the cause, but it is not exactly a Scooby-Do mystery...
;-)
Hope this helps...
Keith Moore wrote:
Was it on AIX?
We have seen similar strange things there. I don't remember all the details,
but at one point there were no background processes, yet we could still logon
using sqlplus! Couldn't do anything, but it said we were connected.
When we tried to startup the database, it also told us that it was already
running.
Keith
Ok, I just ran into a strange one this morning. US central time that is.
We are moving our clustered databases to new servers, and we put one
database onto the new hardware for initial testing. We restored from a
backup, and got both instances running. Today when I did a
ps -ef | grep pmon
I saw two pmon processes going for the single instance. I have no idea what
could have caused that. So I set my oracle_sid, path, oracle_Home and
shutdown the instance. What do you know, there was still one pmon process
running. I did a kill -9 on that one.
Then I go to start the instance again. And I get this error: ORA 01081
"cannot start already-running ORACLE - shut it down first"
I did a ps -ef | grep oracle, didnt get anything (except my user process of
course).
I finally googled and found one possibility, a locked shared memory segment.
Sure enough
ipcs -a | grep dba
returned a large segment of memory that was still locked. So I ran this
(found on the same site) and released it:
ipcs -a | grep dba | perl -ane 'system "ipcrm -$F[0] $F[1]"'
My question is, has anyone seen this before, or know what causes it?
--
Andrew W. Kerber
'If at first you dont succeed, dont take up skydiving.'
--
http://www.freelists.org/webpage/oracle-l
--
http://www.freelists.org/webpage/oracle-l
Received on Wed Sep 26 2007 - 13:56:03 CDT
