Last night we thought we'd do the 'quick' SP3 to SP4a update. SP3 had all post-sp3 patches applied. Proliant ML350 3.5 Gig RAM, single processor, hyperthreading disabled, all tfs volumes, plenty of free storage space. All 3rd-party programs unloaded, 'java -exit' prior to starting.

All went fine until it got to around 78% and then it appeared to hang. I have had enough upgrade experience with upgrades that I knew to leave it alone - some processes can take up to an hour for some unfixed reason. Received an error message after about 1/2 hour on the install screen:

sys:java\lib\ext\mysql-connector\java\*.jar cannot be removed. Error Code:142 (8e hex) ICMD-5.0-13 (something close to that) .

We hit <Enter> to continue and it just dropped back to the nwconfig screen. We thought - well, how bad can that be - just can't remove a file....

Upon reboot, no JAVA was available, so it kept giving Processor Exception errors. We would run server -na and then type the autoexec entries one at a time - no exceptions - but java still wasn't there. However, just bringing it up with 'server' would give the exception errors.

We looked the error up and found a TID from 2004 that talked about that problem with 'tomcat' instead of this file during an upgrade. Said to try again and watch carefully for when it attempts to inflate the, switch to the console screen, and type java -exit. Well, okay, we'll give it another shot, expecially since we knew JAVA wasn't even there, so we wouldn't have to type java -exit. We did that, watched for - and of course java -exit gave the unknown message since java.nlm didn't seem to exist anymore... It went by that...

Well, again at about 78% it appeared to hang trying to extract - the console screen showed it inflating I don't remember what exactly happened then - we were rather frustrated and I took poor notes, but we got some error after patiently waiting for it to fail.

Well, okay, good thing we chose to backup our original SP3 files when asked in the beginning. So, we got the server up to where we could get into nwconfig and point to the sys:\system\backspx directory. Seemed to work great. we rebooted the server. Uh oh, things have gotten worse now. Very colorful screen showing multiple processor errors and a complete system freeze.

We have had 'down server upgrade' experience from our last 6.0 SP5 to 6.5 SP3 upgrade (java issues I believe), so we ended up doing that. We had to go back to the original SP3 version since I hadn't made/brought SP4a overlays - thinking just the straight SP4a update would work just fine. I did have the SP3 overlays though. All worked and we're now right back where we started.

Just thought I'd share our experiences. I have read threads stating where people have had no problems with this upgrade - we just happened to hit them. We have had very good experience going from NW 6.0 SP5 to OES SP1, using the overlays, on about 15 of our 55 servers so far. I have also done the manual SP4a updates to our SP4 servers with no problems - still get the 'logical address space is running low' messages on them though...

I was wondering, since the 'down server upgrade' has worked for us every time after a failure, could we actually just do a down server upgrade from NW6.5 SP3 to NW6.5 SP4a using the 4a overlay files right from the start - manually creating the c:\nwupdate\health.log file first? We are quite hesitant to try this straight SP3 to SP4a update again on our few other servers that require it. On our remaining NW6.0 SP5 servers we'll use the SP4a overlays, since from the threads they seem more crash-proof.