Hi Uwe,
With this you can patch the flag (rstConfigPhase_kb00 to none=0) located in the restartrecord page on the system devspace.Today I had a power outage, and one of my databases produced a -9212 (db_stop, which helps sometimes didn't help in that case). I've uploaded tons of data over the weekend, and didn't get to perform a backup.
I'm trying to use x-diagnose online diagnosis to force the database into a "consistent state"; when I'm trying 'edit restart record' (instance is in cold mode), I can use INSERT to change the values, but I cannot save the changes. How do I achieve this?
rescue my data!!! pleeze! ;)
regards
Toni
-----Original Message-----_______________________________________________
From: Watz [mailto:[EMAIL PROTECTED]]
Sent: Samstag, 30. November 2002 02:02
To: [EMAIL PROTECTED]
Subject: successful -9212 kernel hack to get db up warm again
Hi,
a rather large database on my notebook wouldn't go warm anymore with the -9212 error "previous restart incomplete" after shutting down Windows without stopping the database explicitely (which works 99.9% of the time without troubles...I read its a possible rare bug in older SAPDB releases somewhere).
Well typically the only advise you get on this error is to restore a backup because the database is said to be "inconsistent". This doesn't help alot however if you don't have a backup at all and just would like to access whatever data is left...especially if its the work of several days.
What I did now was upgrading the 7.2.4.3 Database to the latest 7.3.0.29 version.
It still wouldn't start up, it obviously saved some flag somewhere so it wouldn't even try to restart again and rather fail right away instead.
I managed to hack a few parts of the kernels ugly pascal sources so it would ignore the state it saved in the "restart page", and it indeed worked !
I got my database up and running again, shut it down cleanly and exchanged the hacked kernel by the original one again.
It works fine now, and a "database check" didn't give any errors (however I have no idea how reliable that is). Since the database is running in logmode DEMO anyway I shouldn't have troubles with any bad or not loaded redo log pages as a result of the hack I guess. At least I have my data back and can copy it over into a safe database.
What about implementing such a "last resort" switch officially (come on, you don't have to mention it to R/3 customers :-)? I didn't see anything in the code itself, and I'm not sure where the "restart page" is actually saved (sys devspace?) so it could possibly be accessed from outside.
I mean if theres no backup to restore and the DB cannot go warm by any means, what would one have to loose?
So well for any people having the same problem in the future....I attached the two hacked files (vkb81/vkb82) that can be used to build a 7.3.0.29 recovery kernel like I used it. If you have luck it may work for your case, too.
It may be your last resort if you have no backup and no other way to get the db up warm again due to the -9212 error.
I'm just glad that didn't happen with Oracle or other closed-source DBs to me.
I would have had no way to get around this problem and no quick and cheap way access my data.
Watz
sapdb.general mailing list
[EMAIL PROTECTED]
http://listserv.sap.com/mailman/listinfo/sapdb.general
-- Anton Epple Genomatix Software GmbH Landsberger Strasse 6 D-80339 Muenchen http://www.genomatix.de
