From: Jenny Butler (jbutler@utmem.edu)
Date: Fri Feb 21 2003 - 10:05:57 EST
I sent a message to the list on January 28, 2003 concerning a false
error
with the Oracle 9i database check script on a two-node Tru64 Unix v5.1A
and TruCluster v5.1A configuration. I was using a copy of the example
check script that comes with the TruCluster kit, which basically uses the
ps -e -o pid,command to get a list of processes and look for the one you
specify. In this case I was looking for, among others, the ora_arc0_tst,
ora_ckpt_tst, ora_dbw0_tst, and a number of others.
I finally put in the check script a section to send the entire ps
command
output to a log file and here is what I found. For some reason, in the log
that
was created about 4:40 am today, the output was:
1374341 [oracle]
for the pass looking for ora_arc0_tst but in later passes where it was
looking
for the other processes, the output was:
1374341 ora_arc0_tst
This explains why I got a false error condition. The process did not go
away,
but the information passed back to the ps command did. I also changed the
returns so that it would never return an error, just for testing.
I read the ps man pages and found that the -f would:
-f Attempts to generate a full listing. Under certain circumstances, the
-f option does not return the full path and arguments of a process.
Instead, it will return a bracketed process name such as
[process_name]. This behavior conforms with the XPG4 standard.
Of course, I am not using the -f on the command.
I have not decided how to overcome this problem, but maybe someone
else can offer suggestions. Also, over time someone else using Oracle may
experience this and be helped. I will also send the log and this
information to
the TruCluster support folks for their evaluation.
Thanks, Jenny Butler
jbutler@utmem.edu
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:08 EDT