Questions about applying 8.1.7.2 patchset; Strange memory behavior on unix server.

Hello all,

I am about to apply patchset 8.1.7.2 this Sunday. I have a 24x7 db that is extremely difficult to get downtime for so I have some questions before applying the patch.

This is what I have so far and would like to know if I am correct in applying these or not.

1. patchset 8.1.7.2
2. patch for #1888560 (utlrcmp.sql not present in 8.1.7.2)
3. patch for #1849689 (spatial 8.1.7.2 patchset 1)
4. patchset #1776437 (this was originally for 8.1.7.1 but I read that it's not included in the 8.1.7.2 patchset so I need verification on whether or not it needs to be ran.)

These are all the patches that I see that I was able to download without a password, but there seems to be an awful lot of one-off patches for 8.1.7.2 that require passwords to access. So my main question is: ARE there any other patches that I need to apply this coming Sunday as it will be quite some time before I will be able to get downtime again should other problems arise from applying these patches!

System Information is as follows:

Sun E420r running Solaris 8 with system disks mirrored via Veritas and a T3 array setup with Veritas Volume manger running vxfs where the Oracle software and database reside.
Oracle Database version 8.1.7.1
The last patch applied 8.1.7.1b and I have the parameter _db_handles_cached = 0 turned on at present as the workaround to the SGA memory leak bug which 8.1.7.2 is supposed to fix.)

In addition to the questions posed above, I have noticed some strange behavior on my E420r unix server. Once the database is started, the freemem and freeswap on the server increasingly decreases over time. I only noticed this as it caused a problem several days ago that was giving my users the oracle error "TNS could not start a dedicated server process" and they could not log into the database and even after clearing all connections the problem remained and I had to actually reboot the server to clear the issue! So I started watching the freemem and freeswap and have steadily seen it decline over the past week. I have 4gig memory on the box and the Oracle SGA is set at 1.7gig at present and I do not run MTS so each user gets a dedicated server process and I can have hundreds of these at any one time that are inactive in the database but still consuming resources at the unix level. I initially thought it could be related to the amount of users logging into the database (there are no direct user connections to the server itself only thru Oracle) but after monitoring it keeps declining wether there are 300 users or wheter there are 800 users which leans toward the behavior being an actual memory leak on the server as one would assume that as users logoff the database the system would get it's resources back but it's just the opposite as it just keeps declining?? Very strange, Has anyone else reported or seen this kind of behavior on their server(s)???

Thanks in advance for any input!!

Chad Smith
DBA
Industrial Light & Magic
415-448-3062 Received on Fri Aug 24 2001 - 13:52:03 CDT