I did an patch upgrade to 10G R1 (10.1.0.2) -> 10.1.0.4 on linux last night.
The patch went thru fine, cygwin x-window connected to linux with xauth.
A couple of clicks and then simple @?/rdbms/admin/catpatch.sql from sqlplus and its done.
However we have been having issues with Oracle being able to see more than 3 Gig of memory on linux. There is a solution, however I haven't seen the solution. I will shortly...
I restart the db with a older parameter file, save that as spfile and do the catpatch.
However the alert.log complained about the parameter PARALLEL_AUTOMATIC_TUNING being depreciated. I remove the parameter and restart.
Later on the production support blokes can't connect, they are getting TNS-12516, TNS-12560 errors. I log on from home can tnsping, can connect locally.
Eventually even sqlplus "/ as sysdba" as oracle from unix doesn't connect (without any errors as to why). Finally Oracle throws an max processes exceeded error when I try to connect as a different user.
It turns out setting the depreciated parameter PARALLEL_AUTOMATIC_TUNING = false (or not specifying it) means Oracle sets the processes to 30 automatically!!!!
I found that info in a forum entry on metalink (Oracle support website). Apparently that behaviour is undocumented and was raised to be documented. It is not there yet as I use the Oracle documentation libraries from tahiti.oracle.com (requires technet login)
I fixed set the PROCESSES parameter to a reasonable number and the connection isssues (and TNS errors) disappear.
Have Fun
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment